feat_line_text
stringlengths 4.48k
87.5k
| feat_filename_text
int64 9
32.7k
| text
stringlengths 4.21k
81.5k
| feat_shorter_abstract
stringlengths 218
3.41k
| target
stringlengths 711
3.65k
| evaluation_predictions
sequence |
|---|---|---|---|---|---|
BACKGROUND
metagenomics is a relatively recent field of research, dealing primarily with the investigation of microbial consortia of uncultivable organisms.
it has enabled the study of microbiota sampled directly from environmental niches, such as the ocean <cit> , soil <cit> , hot springs <cit> , hydrothermal vents <cit> , polar ice caps <cit> and hypersaline environments <cit> .
in depth investigation of these consortia have given insight into microbial ecology <cit> , diversity <cit> , as well as archeal lineages <cit> .
not only is such knowledge valuable to the understanding of our biosphere, it has also facillatated advancement of biotechnology <cit> , human physiology <cit> and sequencing of contaminated samples of now extinct species <cit> , to name a few.
prior to the metagenomics approach, we see that attempts at characterising microbial communities using pure clonal samples of constiuent organisms resulted in a low discovery rate of novel microbes <cit> .
metagenomics is able to tackle this problem by means of direct sequencing of an environmental sample without the need for a pre-cloning step.
this enables approximately 99% of earth's undiscovered microbiota, which resist standard laboratory culturing techniques, to be sequenced and analysed.
however, when an environmental sample is sequenced en masse, a fundamental computational challenge is introduced: the characterisation of sequenced reads with respect to their phylogenetic origin <cit> .
such in silico profiling of sequenced dna is referred to as binning.
binning is an important first step to further downstream analysis of a metagenome.
of particular interest in this preliminary stage of analysis is the taxonomic composition of the sample, and further, the association between sequenced dna fragments and their parent organism.
many reported attempts at this analysis are founded on one of three key concepts: marker gene based assignment, sequence similarity assignments or sequence composition based assignments.
taxonomic profiling using conserved marker genes
through various stages of an organism's evolution changes take place in the composition its genome, permitting adaptation to changes in the environment, for example.
different locations in the genome experience distinct rates of change.
hyper-variant regions are particularly found in non-coding, inter-genic regions <cit> .
this is because pronounced changes in genes that code for particular functions will degrade characteristic functionality of an organism.
the exceptions to this are slowly evolving marker genes in the guise of non-coding ribsomal rnas.
these conserved marker genes have been fundamental to the study of microbial phylogeny <cit> .
prior to the discovery of such marker genes, phylogenetic analysis of microbes revealed the existence of only two primeval lineages.
however, a founding study <cit> highlighted the insuffciency of existing approaches to capture all extant lineages.
it was this which lead to the establishment of three primary kingdoms or domains of archea, bacteria and eucarya.
more recently, studies that adopt this approach have greatly contributed to our knowledge of the actual diversity of microorganisms <cit> .
automatation of the phylotyping procedure for metagenomic dna using 16s rrna markers are also becoming more prevalent <cit> .
in <cit> two fundamental questions are posed regarding the number of bacterial phylotypes that can co-exist and the way in which they are organised, and an attempt at addressing both has been through 16s rrna sampling of a bacterioplankton assemblage.
as <cit> argues, these two pieces of information are critical to the understanding of function, population biology and biogeography.
further, <cit> uses 16s rrna libraries to compare the phylogenetic distance between various microbial communities.
on the one hand where such marker genes are providing valuable insight into the composition of microbial assemblages, they carry with them the inability to characterise the majority of sequenced reads.
it has been reported that these marker genes appear infrequently in a typical set of sequenced reads, despite the high density of open reading frames found in microbial genomes <cit> .
consequently, only a fraction of reads can be assigned.
read length is also a factor to consider when attempting classification using marker genes.
16s rrna gene are generally <dig> base pairs in length and as such will be distributed over multiple reads of current state-of-the-art sequencing technologies.
complications then arise in taxonomic profiling when marker genes are assumed to be partially located on sequenced reads.
however, this is not so much a critical issue for future sequencing platforms, as whole-molecule sequencing is designed to deliver sequenced reads that are in the order of length of the marker genes <cit> .
marker genes will continue to be used for taxonomic characterisation of a metagenome, as it is arguably the most accurate <cit> .
characterisation based on similarity to known sequences
a viable alternative to the use of conserved regions of genomic dna is the use of previously sequenced homologs as a basis for phylogenetic characterisation.
databases of complete genomes and information on genes, protein families and so forth have experienced exponential growth, especially over the last few decades, attributed in part to advances made by marker gene based profiling of microbial organisms.
tapping into this resource pool has been the focus of a variety of methods.
for instance, early similarity methods used a simple blast search against databases of previously sequenced complete genomes to assign short fragments to a particular taxonomic rank, using homologs <cit> or orthologs <cit> .
currently, the length of reads is an important challenge <cit> and methods that are able to assign short stretches of dna are required, particularly where next generation sequencing platforms are used to sequence metagenomes.
in fact, claims have even been made to a solution for phlyotyping a metagenome irrespective of the sequencing technique used <cit> .
the strength of these methods relies on the length of sequences able to be assigned.
we see that recent attempts achieve accurate classification of reads down to <dig> base pairs in length <cit> .
despite this, for real metagenomic samples, fewer than 10% of the reads could be identified using the pfam domain and protein families <cit> , which is suggestive of poor database coverage over extant microorganisms.
of this 10%, only a fraction could be assigned to a particular lineage.
in general, a bottleneck exists in conducting the initial search against various databases, often necessitating the use of a large number of cpu hours on high performance computing solutions.
as these databases continue to grow at their current rate, this bottle-neck will increasingly impose significant delays in any metagenomic project, perhaps with minimal pay-off for the classification of novel organisms.
however, by adopting these strategies we are fundamentally relying on the assumption that novel microbes sampled from the environment will be represented in existing databases.
with an abundance of microbial diversity yet to be discovered, it is counter-intuitive to found decisions on previously discovered genomes and protein sequences <cit> , particularly when the majority of these are derived from culture-dependant techniques, while an estimated 99% of novel microbes are yet to be discovered can not be cultured using current in vivo techniques <cit> .
characterisation based on sequence composition
the composition of a dna sequence is defined by the non-random ordering of its base-pairs, in terms of the four atomic nucleotides.
taking into account that there are specific causes and evolutionary factors for the variation in base composition of genomic dna, methods have been developed to extract common patterns between organisms at varying levels of taxonomic resolution, such that sequences of similar species are able to be grouped.
the general trend with compositional approaches has been the modification of machine learning methods to work around existing compositional feature spaces.
deterministic pattern spaces such as oligonucleotide frequency counts are among the more dominant of choices among compositional binning methods.
these operate on the assumption that the relative abundance of certain words - also referred to as oligos, i.e.
acgta is a 5-mer oligo - primarily dictate the association of one sequence to another.
this is particularly useful for observing codon usage biases, for example.
in fact, compositional biases, particularly in the case of tetranucleotide frequencies, is hypothesised to have strong biological significance in terms of phylogeny <cit> .
it is further argued that the larger number of permutations possible in tetranucleotide frequencies allows greater authority to discriminate between genomic fragments from different genomes.
this argument holds if the conditions of low intragenomic variation and large sequence length hold.
due to the large number of combinations of oligos possible, tetranucleotide frequency has been reported to have greater discriminatory power than the g-c content of a sequence, for instance <cit> .
clustering tetranucleotide frequencies using fixed-size self-organising maps has been shown to be possible <cit> .
however, the imposition of a fixed size som has been attributed to feature maps that do not faithfully represent the input data, and as such the growing self-organising map has been used to alleviate this flaw <cit> .
in the context of current next generation sequencing technologies, we find that reads are as short as 80- <dig> base pairs.
in light of this, methods that operate on nucleotide frequency alone are at a disadvantage, as a signal strong enough to make inferences on phylogenetic origin of a sequence requires long stretches of dna.
it has been shown that <dig> bp is an acceptable sequence length to make accurate predictions <cit> , yet it has also been shown that sequences as short as <dig> bp can be classified <cit> .
as yet, the <dig> bp barrier - as it has been colloquially termed in the literature - is still an open challenge.
to counteract this limitation, methods that adopt nucleotide frequencies as a means of sequence representation typically operate on assembled contigs.
however, for complex communities the required amount of coverage for modest assembly translates to a substantial sequencing requirement.
the feasibility, therefore, for current composition based methods looks to be limited to microbial consortia with minimal to moderate diversity.
further, these methods compensate for weak phylogenetic signals by consolidating information from other sources, such as external databases.
methods have also used training data from potential homologs from public databases to construct a representative signal for a particular clade <cit> .
in the same study it is also shown that it is possible to generate training data directly from the metagenome itself, but it is argued that at least <dig> kbp of data is required to construct an accurate model for a particular organism at a particular taxonomic rank.
the literature suggests that significant advances in compositional approaches to the binning problem have primarily looked at the issue of representing the composition of a sequence, rather than refining machine learning methods that operate in a sub-par feature space.
such is the case with the succession of g-c content by tetranucleotide frequencies <cit> , for example.
there is much that can be unveiled when patterns are extracted from dna sequences, and since compositional methods are generally database-independent they are not susceptible to cloning biases observed in similarity-based methods.
it is, however, a matter of knowing what patterns to extract and how best to extract them.
here we propose the use of the oligonucleotide frequency derived error gradient as a feature space for the characterisation of dna sequences from isloate organisms.
the proposed feature space relies on the concept of oligonucleotide frequency profiles and their demonstrated ability to characterise genomic dna.
RESULTS
evaluation on simulated metagenomic data sets
data set description
recently published metagenomic benchmark data sets <cit> have been selected to evaluate the binning performance using our proposed dna sequence representation.
the benchmarks were formed using real dna sequences of <dig> isolate microbial genomes, sequenced at doe joint genome institute.
the dominant strain in the simlc set is given a coverage of approximately <dig> × and just over <dig> mbp in total sequence length.
the dominant strain is anked by <dig> lower abundance strains, with coverage less than <dig> ×.
simmc introduces three dominant strains which are represented with coverages ranging from <dig> × to <dig> ×.
characteristic of agricultural soil, the simhc data set contains no dominant strains, and has poor coverage.
the highest organism coverage in this data set is at an estimated <dig> ×.
evaluation procedure
as previously described, the simulated metagenome data contains three data sets designed to represent three microbial communities of varying degrees of complexity.
here we present the analysis of binning performance on the medium complexity simmc data set, as this serves as a basis for comparison against current compositional binning techniques.
similar to <cit> , we conduct two tests to evaluate the quality of binning using ofdeg.
the first takes only contigs which are greater than <dig> bp in length - as these were deemed to have a minimal degree of chimerism.
the second takes major contigs, which are those that are assembled using at least <dig> reads.
we use both the assembler output of arachne <cit> and phrap for evaluation, where phrap produces shorter contigs but is deemed more reliable <cit> .
contigs generated by jazz were excluded from the analysis <cit> .
similarly, for the purposes of comparison we restrict ourselves to the taxonomic rank of order, using ncbi's taxonomy definition.
for all tests conducted, ofdeg values were computed on the basis of a 4-mer of profile for comparison against tetranucleotide frequency recommended by <cit> .
the sampling depth was set to <dig> and a step size of 10% of the sequence length was used.
each ofdeg value for a sequence is an average over <dig> - determined empirically - subsequences which were truncated to the length of the shortest contig for each test, given the criteria of selecting contigs as defined above.
all the ofdeg values computed used 80% of the sequence length.
in an attempt to increase the discriminatory power of ofdeg for sequence separation, we also consider the effect of ofdeg in conjunction with g-c content as it has been used previously to successfully characterise organisms and maintains the low dimensionality of the proposed feature space.
measuring the accuracy
accuracy is measured by how well a particular organism is characterised.
with respect to each organism, the bin that is identified as containing the maximum number of fragments for that particular organism is considered as its reference bin.
fragments assigned to the reference bin that are of the reference organism are deemed the true positives.
similarly, fragments contained in the reference bin that are not of the reference organism are false positives.
fragments of the reference organism that are located outside the reference bin are considered false negatives and lastly, fragments that are not of the reference organism and located outside the reference bin are considered true negatives.
using these interpretations we use the standard definition of sensitivity and specificity to evaluate the quality of binning.
for the semi-supervised case we look at the label assigned to each fragment and compare this assignment to its true origin.
for each class of organism, at a specific taxanomic rank, we look at how many of those fragments have been assigned correctly using the same definitions as for the unsupervised case.
fragments that are not assigned a label are not included in the calculation.
unsupervised setting
we first evaluate the fidelity of ofdeg by clustering computed ofdeg values in an unsupervised manner.
for compositional methods, the relative abundance of oligonucleotide frequencies and inherent biases therein have been linked to a sequence's phylogenetic origin.
raw ofdeg values are clustered using partitioning around mediods <cit> , and the average silhouette width is used to compute the most representative number of classes based on the clustering structure.
in general, it is clearly evident that clustered ofdeg values in conjunction with g-c content improves on the performance of tetranucleotide frequency alone .
the lower sensitivity value for tetranucleotide frequency can be explained by the less defined clusters that result, where fragments that are associated with one genome type are distributed across multiple bins.
as the reference sequence length for ofdeg calculation must be the same for all sequences in order to produce meaningful comparisons, the minimum contig lengths in the <dig> reads tests highlighted the benefits of using ofdeg.
for the phrap assembled data, the minimum contig length is only <dig> bp, yet we see that the binning fidelity is competitive with features that require the entire sequence length.
with an increase in minimum contig length for the arachne assembled data to <dig> bp, again for the <dig> reads test, the binning performance increases.
this is even more prevalent when considering the <dig> kbp tests, which have a minimum contig length of <dig> kbp, where a near perfect separation of the two out of three dominant species is observed.
semi-supervised setting
in this setting we require the use of a minimal amount of annotated data which we will refer to as labels, a basis for supervised learning.
using labels, we are able to investigate the possibility of a performance increase using existing knowledge of sequenced microbial dna, an observation described in <cit> .
here we also use the s-gsom algorithm <cit> .
fixed-sized soms were not used, as the imposition of a fixed sized map may result in incorrect representation of the input space, and subsequently poor clustering.
note, this was used only as a comparison of features, and not a contribution as a novel clustering technique.
similar to <cit> we select as labels <dig> kbp flanking regions of the conserved 16s rrna gene, subject to rules also defined therein.
ofdeg values are computed for each flanking sequence of 16s rrna genes found in the genomes used in the simulated data sets.
these values were augmented with the processed data sets as seeds.
additionally, topology type, similarity measure, weight initialisation type, neighbourhood kernel, initial learning rates, and training epochs for s-gsom were selected to be the same.
for comparison, we also evaluate classification performance at cluster percentages of 55% and 75%, as these parameter settings are recommended in <cit> .
particularly for the <dig> kbp tests, we see an improvement in performance over a purely unsupervised attempt at characterising fragments, achieving a sensitivity and specificity of <dig> for both assemblers .
as is appreciated in purely clustered data, there will be ambiguity in assignment at the edges of clusters, i.e.
does fragment x, which lies directly in between the centres of clusters a and b, belong to cluster a or b?
in selecting only those sequences that will give confident predictions, avoiding the previously mentioned situation, the accuracy of binning is increased.
in this case, the number of fragments assigned are reduced as a result.
tests using ofdeg for semi-supervised classification at higher cp values showed a decrease in accuracy, as spurious assignments were made.
the increase in performance experienced by s-gsom using tetranucelotide frequency, on the other hand, can be explained by seed location within a fully trained map.
to be able to use semi-supervised learning, we require an indication of the diversity of the metagenome, perhaps through targeted sequencing of the 16s ribosomal rna.
given this, the flanking sequences are used as reference classes for binning.
however, reference genomes would unlikely be available, given the current knowledge of microbial diversity.
the accuracy is then determined by the correct identification of the taxa that is actually present in the data set.
this, however, is not a requirement for binning, but merely a means to get a sense of species composition.
binning is still able to proceed without such a priori knowledge.
overall perfomance and discussion
in order to capture the overall relative performance of both ofdeg and tf feature spaces, we compute the discriminant ability of each using averaged sensitivity and specificity values over all tests conducted.
with reference to we see that both the unsupervised and semi-supervised methods which use ofdeg+gc as a feature space perform best overall with respect to the four different simmc tests.
though the semi-supervised method outperforms the unsupervised method, the average number of assignments made by the unsupervised variant is far greater.
if labels are available, we are able to classify fragments with approximately 96% accuracy.
however, in the case where labels are not present, unsupervised methods can be applied using sequence ofdeg values with 95% accuracy, but at the same time with higher coverage: <dig> % of sequences in the data set as opposed to <dig> % for the semi-supervised case.
computed ofdeg values using 5-mer oligos for <dig> microbial genomes selected arbitrarily from ncbi.
the values listed here represent the sample mean plus or minus the sample standard deviation, averaged over <dig> sampling sites over the entire length of the genome.
the differences reflect the strength of the signal in relation to sequence length.
an important consideration is data dimensionality.
for instance, projecting a sequence into tf space results in the generation of input vectors of dimension <dig> for each sequence in the data set.
with a large number of sequences, the computational cost of clustering such data will be taxing, even more so with higher order oligos.
alternatively, ofdeg is a consistent, one-dimensional feature space - two at most, when used in conjunction with g-c content - irrespective of the underlying word size, which will be beneficial in the initial analysis of metagenome data.
we appreciate that seeding posseses the capacity for a more accurate classification of environmental dna fragments.
however, care should be taken when using this approach.
as noted earlier, this setting relies on the premis that the 16s rrna flanking sequences are representative of the biota in the sample and are available for each species in the metagenome.
using seeds which are incorrect will degrade the classification performance, even if the clustered data is structured correctly.
practical applications of using ofdeg should take into consideration the following.
current next-generation sequencing technologies are producing output at higher rates and shorter read lengths.
the method proposed operates under the assumption that some assembly has been carried out on the raw sequencer output.
akin to compositional methods, sufficient sequence length is required to make inferences based only on the ordering of base-pairs.
however, this compositional feature appears to make a stronger association between phylogeny and sequence composition given shorter strecthes of dna.
of additional importance are repeated regions in genomic dna.
these will be captured and reflected in the computed ofdeg values, which will be lower in comparison to other seqeuneces.
for the detection and removal of, say, redundant repeats, a preprocessing tool could be used to remove these prior to ofdeg computation.
this, however, is beyond the scope of this work.
we emphasise that this work serves to describe the observation of a characteristic linear gradient and its potential application.
although we are unable to fully explain its theoretical underpinnings or provide an in-depth biological interpretation, we are empirically able to show that it does have links to microbial phylogeny.
CONCLUSIONS
here we have presented a novel representation of short dna sequences, derived from oligonulceotide frequency profiles, which allows for the phylogenetic characterisation of relatively short sequences on the basis of sequence compostion alone.
although we have found that microbial phylogeny is potentially captured in ofdeg, we aim to develop a theoretical framework as well as ellicit its biological meaning.
unsupervised clustering revealed ofdeg related features performed better than standard tetranucleotide frequency in representing a relevant organism specific signal.
the extension to a semi-supervised paradigm again demonstrated an improvement in binning performance when using ofdeg values.
in light of the issues faced with semi-supervised classification of ofdeg values, an interesting prospect for future work is the analysis of seed selection and its influence on the accuracy of fragment classification, especially for data sets which contain short contigs.
expressing each sequence in an appropriate feature space is more beneficial than developing intricate machine learning methods that wrap around feature representations that do not capture phylogenetic signals in short sequences.
there is a pressing need to break away from reliance on assemblers that were designed to handle single genomes, built without consideration for processing significantly different, heterogeneous metagenome sequence data.
addressing the fundamental issue of a suitable representation for short dna sequences has shown potential as a first step toward unbiased, database-independent characterisation of metagenome data and novel microbiota.
| 26,997 | metagenomics is a relatively recent field of research, dealing primarily with the investigation of microbial consortia of uncultivable organisms. it has enabled the study of microbiota sampled directly from environmental niches, such as the ocean, soil, hot springs, hydrothermal vents, polar ice caps and hypersaline environments. in depth investigation of these consortia have given insight into microbial ecology, diversity, as well as archeal lineages. not only is such knowledge valuable to the understanding of our biosphere, it has also facillatated advancement of biotechnology, human physiology and sequencing of contaminated samples of now extinct species, to name a few. prior to the metagenomics approach, we see that attempts at characterising microbial communities using pure clonal samples of constiuent organisms resulted in a low discovery rate of novel microbes. metagenomics is able to tackle this problem by means of direct sequencing of an environmental sample without the need for a pre-cloning step. this enables approximately 99% of earth's undiscovered microbiota, which resist standard laboratory culturing techniques, to be sequenced and analysed. however, when an environmental sample is sequenced en masse, a fundamental computational challenge is introduced: the characterisation of sequenced reads with respect to their phylogenetic origin. such in silico profiling of sequenced dna is referred to as binning. binning is an important first step to further downstream analysis of a metagenome. of particular interest in this preliminary stage of analysis is the taxonomic composition of the sample, and further, the association between sequenced dna fragments and their parent organism. many reported attempts at this analysis are founded on one of three key concepts: marker gene based assignment, sequence similarity assignments or sequence composition based assignments. taxonomic profiling using conserved marker genes through various stages of an organism's evolution changes take place in the composition its genome, permitting adaptation to changes in the environment, for example. different locations in the genome experience distinct rates of change. hyper-variant regions are particularly found in non-coding, inter-genic regions. this is because pronounced changes in genes that code for particular functions will degrade characteristic functionality of an organism. the exceptions to this are slowly evolving marker genes in the guise of non-coding ribsomal rnas. these conserved marker genes have been fundamental to the study of microbial phylogeny. prior to the discovery of such marker genes, phylogenetic analysis of microbes revealed the existence of only two primeval lineages. however, a founding study highlighted the insuffciency of existing approaches to capture all extant lineages. it was this which lead to the establishment of three primary kingdoms or domains of archea, bacteria and eucarya. more recently, studies that adopt this approach have greatly contributed to our knowledge of the actual diversity of microorganisms. automatation of the phylotyping procedure for metagenomic dna using 16s rrna markers are also becoming more prevalent. in two fundamental questions are posed regarding the number of bacterial phylotypes that can co-exist and the way in which they are organised, and an attempt at addressing both has been through 16s rrna sampling of a bacterioplankton assemblage. as argues, these two pieces of information are critical to the understanding of function, population biology and biogeography. further, uses 16s rrna libraries to compare the phylogenetic distance between various microbial communities. on the one hand where such marker genes are providing valuable insight into the composition of microbial assemblages, they carry with them the inability to characterise the majority of sequenced reads. it has been reported that these marker genes appear infrequently in a typical set of sequenced reads, despite the high density of open reading frames found in microbial genomes. consequently, only a fraction of reads can be assigned. read length is also a factor to consider when attempting classification using marker genes. 16s rrna gene are generally base pairs in length and as such will be distributed over multiple reads of current state-of-the-art sequencing technologies. complications then arise in taxonomic profiling when marker genes are assumed to be partially located on sequenced reads. however, this is not so much a critical issue for future sequencing platforms, as whole-molecule sequencing is designed to deliver sequenced reads that are in the order of length of the marker genes. marker genes will continue to be used for taxonomic characterisation of a metagenome, as it is arguably the most accurate. characterisation based on similarity to known sequences a viable alternative to the use of conserved regions of genomic dna is the use of previously sequenced homologs as a basis for phylogenetic characterisation. databases of complete genomes and information on genes, protein families and so forth have experienced exponential growth, especially over the last few decades, attributed in part to advances made by marker gene based profiling of microbial organisms. tapping into this resource pool has been the focus of a variety of methods. for instance, early similarity methods used a simple blast search against databases of previously sequenced complete genomes to assign short fragments to a particular taxonomic rank, using homologs or orthologs. currently, the length of reads is an important challenge and methods that are able to assign short stretches of dna are required, particularly where next generation sequencing platforms are used to sequence metagenomes. in fact, claims have even been made to a solution for phlyotyping a metagenome irrespective of the sequencing technique used. the strength of these methods relies on the length of sequences able to be assigned. we see that recent attempts achieve accurate classification of reads down to base pairs in length. despite this, for real metagenomic samples, fewer than 10% of the reads could be identified using the pfam domain and protein families, which is suggestive of poor database coverage over extant microorganisms. of this 10%, only a fraction could be assigned to a particular lineage. in general, a bottleneck exists in conducting the initial search against various databases, often necessitating the use of a large number of cpu hours on high performance computing solutions. as these databases continue to grow at their current rate, this bottle-neck will increasingly impose significant delays in any metagenomic project, perhaps with minimal pay-off for the classification of novel organisms. however, by adopting these strategies we are fundamentally relying on the assumption that novel microbes sampled from the environment will be represented in existing databases. with an abundance of microbial diversity yet to be discovered, it is counter-intuitive to found decisions on previously discovered genomes and protein sequences, particularly when the majority of these are derived from culture-dependant techniques, while an estimated 99% of novel microbes are yet to be discovered can not be cultured using current in vivo techniques. characterisation based on sequence composition the composition of a dna sequence is defined by the non-random ordering of its base-pairs, in terms of the four atomic nucleotides. taking into account that there are specific causes and evolutionary factors for the variation in base composition of genomic dna, methods have been developed to extract common patterns between organisms at varying levels of taxonomic resolution, such that sequences of similar species are able to be grouped. the general trend with compositional approaches has been the modification of machine learning methods to work around existing compositional feature spaces. deterministic pattern spaces such as oligonucleotide frequency counts are among the more dominant of choices among compositional binning methods. these operate on the assumption that the relative abundance of certain words - also referred to as oligos, i.e. acgta is a 5-mer oligo - primarily dictate the association of one sequence to another. this is particularly useful for observing codon usage biases, for example. in fact, compositional biases, particularly in the case of tetranucleotide frequencies, is hypothesised to have strong biological significance in terms of phylogeny. it is further argued that the larger number of permutations possible in tetranucleotide frequencies allows greater authority to discriminate between genomic fragments from different genomes. this argument holds if the conditions of low intragenomic variation and large sequence length hold. due to the large number of combinations of oligos possible, tetranucleotide frequency has been reported to have greater discriminatory power than the g-c content of a sequence, for instance. clustering tetranucleotide frequencies using fixed-size self-organising maps has been shown to be possible. however, the imposition of a fixed size som has been attributed to feature maps that do not faithfully represent the input data, and as such the growing self-organising map has been used to alleviate this flaw. in the context of current next generation sequencing technologies, we find that reads are as short as 80- base pairs. in light of this, methods that operate on nucleotide frequency alone are at a disadvantage, as a signal strong enough to make inferences on phylogenetic origin of a sequence requires long stretches of dna. it has been shown that bp is an acceptable sequence length to make accurate predictions, yet it has also been shown that sequences as short as bp can be classified. as yet, the bp barrier - as it has been colloquially termed in the literature - is still an open challenge. to counteract this limitation, methods that adopt nucleotide frequencies as a means of sequence representation typically operate on assembled contigs. however, for complex communities the required amount of coverage for modest assembly translates to a substantial sequencing requirement. the feasibility, therefore, for current composition based methods looks to be limited to microbial consortia with minimal to moderate diversity. further, these methods compensate for weak phylogenetic signals by consolidating information from other sources, such as external databases. methods have also used training data from potential homologs from public databases to construct a representative signal for a particular clade. in the same study it is also shown that it is possible to generate training data directly from the metagenome itself, but it is argued that at least kbp of data is required to construct an accurate model for a particular organism at a particular taxonomic rank. the literature suggests that significant advances in compositional approaches to the binning problem have primarily looked at the issue of representing the composition of a sequence, rather than refining machine learning methods that operate in a sub-par feature space. such is the case with the succession of g-c content by tetranucleotide frequencies, for example. there is much that can be unveiled when patterns are extracted from dna sequences, and since compositional methods are generally database-independent they are not susceptible to cloning biases observed in similarity-based methods. it is, however, a matter of knowing what patterns to extract and how best to extract them. here we propose the use of the oligonucleotide frequency derived error gradient as a feature space for the characterisation of dna sequences from isloate organisms. the proposed feature space relies on the concept of oligonucleotide frequency profiles and their demonstrated ability to characterise genomic dna. evaluation on simulated metagenomic data sets data set description recently published metagenomic benchmark data sets have been selected to evaluate the binning performance using our proposed dna sequence representation. the benchmarks were formed using real dna sequences of isolate microbial genomes, sequenced at doe joint genome institute. the dominant strain in the simlc set is given a coverage of approximately × and just over mbp in total sequence length. the dominant strain is anked by lower abundance strains, with coverage less than ×. simmc introduces three dominant strains which are represented with coverages ranging from × to ×. characteristic of agricultural soil, the simhc data set contains no dominant strains, and has poor coverage. the highest organism coverage in this data set is at an estimated ×. evaluation procedure as previously described, the simulated metagenome data contains three data sets designed to represent three microbial communities of varying degrees of complexity. here we present the analysis of binning performance on the medium complexity simmc data set, as this serves as a basis for comparison against current compositional binning techniques. similar to, we conduct two tests to evaluate the quality of binning using ofdeg. the first takes only contigs which are greater than bp in length - as these were deemed to have a minimal degree of chimerism. the second takes major contigs, which are those that are assembled using at least reads. we use both the assembler output of arachne and phrap for evaluation, where phrap produces shorter contigs but is deemed more reliable. contigs generated by jazz were excluded from the analysis. similarly, for the purposes of comparison we restrict ourselves to the taxonomic rank of order, using ncbi's taxonomy definition. for all tests conducted, ofdeg values were computed on the basis of a 4-mer of profile for comparison against tetranucleotide frequency recommended by. the sampling depth was set to and a step size of 10% of the sequence length was used. each ofdeg value for a sequence is an average over - determined empirically - subsequences which were truncated to the length of the shortest contig for each test, given the criteria of selecting contigs as defined above. all the ofdeg values computed used 80% of the sequence length. in an attempt to increase the discriminatory power of ofdeg for sequence separation, we also consider the effect of ofdeg in conjunction with g-c content as it has been used previously to successfully characterise organisms and maintains the low dimensionality of the proposed feature space. measuring the accuracy accuracy is measured by how well a particular organism is characterised. with respect to each organism, the bin that is identified as containing the maximum number of fragments for that particular organism is considered as its reference bin. fragments assigned to the reference bin that are of the reference organism are deemed the true positives. similarly, fragments contained in the reference bin that are not of the reference organism are false positives. fragments of the reference organism that are located outside the reference bin are considered false negatives and lastly, fragments that are not of the reference organism and located outside the reference bin are considered true negatives. using these interpretations we use the standard definition of sensitivity and specificity to evaluate the quality of binning. for the semi-supervised case we look at the label assigned to each fragment and compare this assignment to its true origin. for each class of organism, at a specific taxanomic rank, we look at how many of those fragments have been assigned correctly using the same definitions as for the unsupervised case. fragments that are not assigned a label are not included in the calculation. unsupervised setting we first evaluate the fidelity of ofdeg by clustering computed ofdeg values in an unsupervised manner. for compositional methods, the relative abundance of oligonucleotide frequencies and inherent biases therein have been linked to a sequence's phylogenetic origin. raw ofdeg values are clustered using partitioning around mediods, and the average silhouette width is used to compute the most representative number of classes based on the clustering structure. in general, it is clearly evident that clustered ofdeg values in conjunction with g-c content improves on the performance of tetranucleotide frequency alone. the lower sensitivity value for tetranucleotide frequency can be explained by the less defined clusters that result, where fragments that are associated with one genome type are distributed across multiple bins. as the reference sequence length for ofdeg calculation must be the same for all sequences in order to produce meaningful comparisons, the minimum contig lengths in the reads tests highlighted the benefits of using ofdeg. for the phrap assembled data, the minimum contig length is only bp, yet we see that the binning fidelity is competitive with features that require the entire sequence length. with an increase in minimum contig length for the arachne assembled data to bp, again for the reads test, the binning performance increases. this is even more prevalent when considering the kbp tests, which have a minimum contig length of kbp, where a near perfect separation of the two out of three dominant species is observed. semi-supervised setting in this setting we require the use of a minimal amount of annotated data which we will refer to as labels, a basis for supervised learning. using labels, we are able to investigate the possibility of a performance increase using existing knowledge of sequenced microbial dna, an observation described in. here we also use the s-gsom algorithm. fixed-sized soms were not used, as the imposition of a fixed sized map may result in incorrect representation of the input space, and subsequently poor clustering. note, this was used only as a comparison of features, and not a contribution as a novel clustering technique. similar to we select as labels kbp flanking regions of the conserved 16s rrna gene, subject to rules also defined therein. ofdeg values are computed for each flanking sequence of 16s rrna genes found in the genomes used in the simulated data sets. these values were augmented with the processed data sets as seeds. additionally, topology type, similarity measure, weight initialisation type, neighbourhood kernel, initial learning rates, and training epochs for s-gsom were selected to be the same. for comparison, we also evaluate classification performance at cluster percentages of 55% and 75%, as these parameter settings are recommended in. particularly for the kbp tests, we see an improvement in performance over a purely unsupervised attempt at characterising fragments, achieving a sensitivity and specificity of for both assemblers. as is appreciated in purely clustered data, there will be ambiguity in assignment at the edges of clusters, i.e. does fragment x, which lies directly in between the centres of clusters a and b, belong to cluster a or b? in selecting only those sequences that will give confident predictions, avoiding the previously mentioned situation, the accuracy of binning is increased. in this case, the number of fragments assigned are reduced as a result. tests using ofdeg for semi-supervised classification at higher cp values showed a decrease in accuracy, as spurious assignments were made. the increase in performance experienced by s-gsom using tetranucelotide frequency, on the other hand, can be explained by seed location within a fully trained map. to be able to use semi-supervised learning, we require an indication of the diversity of the metagenome, perhaps through targeted sequencing of the 16s ribosomal rna. given this, the flanking sequences are used as reference classes for binning. however, reference genomes would unlikely be available, given the current knowledge of microbial diversity. the accuracy is then determined by the correct identification of the taxa that is actually present in the data set. this, however, is not a requirement for binning, but merely a means to get a sense of species composition. binning is still able to proceed without such a priori knowledge. overall perfomance and discussion in order to capture the overall relative performance of both ofdeg and tf feature spaces, we compute the discriminant ability of each using averaged sensitivity and specificity values over all tests conducted. with reference to we see that both the unsupervised and semi-supervised methods which use ofdeg+gc as a feature space perform best overall with respect to the four different simmc tests. though the semi-supervised method outperforms the unsupervised method, the average number of assignments made by the unsupervised variant is far greater. if labels are available, we are able to classify fragments with approximately 96% accuracy. however, in the case where labels are not present, unsupervised methods can be applied using sequence ofdeg values with 95% accuracy, but at the same time with higher coverage: % of sequences in the data set as opposed to % for the semi-supervised case. computed ofdeg values using 5-mer oligos for microbial genomes selected arbitrarily from ncbi. the values listed here represent the sample mean plus or minus the sample standard deviation, averaged over sampling sites over the entire length of the genome. the differences reflect the strength of the signal in relation to sequence length. an important consideration is data dimensionality. for instance, projecting a sequence into tf space results in the generation of input vectors of dimension for each sequence in the data set. with a large number of sequences, the computational cost of clustering such data will be taxing, even more so with higher order oligos. alternatively, ofdeg is a consistent, one-dimensional feature space - two at most, when used in conjunction with g-c content - irrespective of the underlying word size, which will be beneficial in the initial analysis of metagenome data. we appreciate that seeding posseses the capacity for a more accurate classification of environmental dna fragments. however, care should be taken when using this approach. as noted earlier, this setting relies on the premis that the 16s rrna flanking sequences are representative of the biota in the sample and are available for each species in the metagenome. using seeds which are incorrect will degrade the classification performance, even if the clustered data is structured correctly. practical applications of using ofdeg should take into consideration the following. current next-generation sequencing technologies are producing output at higher rates and shorter read lengths. the method proposed operates under the assumption that some assembly has been carried out on the raw sequencer output. akin to compositional methods, sufficient sequence length is required to make inferences based only on the ordering of base-pairs. however, this compositional feature appears to make a stronger association between phylogeny and sequence composition given shorter strecthes of dna. of additional importance are repeated regions in genomic dna. these will be captured and reflected in the computed ofdeg values, which will be lower in comparison to other seqeuneces. for the detection and removal of, say, redundant repeats, a preprocessing tool could be used to remove these prior to ofdeg computation. this, however, is beyond the scope of this work. we emphasise that this work serves to describe the observation of a characteristic linear gradient and its potential application. although we are unable to fully explain its theoretical underpinnings or provide an in-depth biological interpretation, we are empirically able to show that it does have links to microbial phylogeny. here we have presented a novel representation of short dna sequences, derived from oligonulceotide frequency profiles, which allows for the phylogenetic characterisation of relatively short sequences on the basis of sequence compostion alone. although we have found that microbial phylogeny is potentially captured in ofdeg, we aim to develop a theoretical framework as well as ellicit its biological meaning. unsupervised clustering revealed ofdeg related features performed better than standard tetranucleotide frequency in representing a relevant organism specific signal. the extension to a semi-supervised paradigm again demonstrated an improvement in binning performance when using ofdeg values. in light of the issues faced with semi-supervised classification of ofdeg values, an interesting prospect for future work is the analysis of seed selection and its influence on the accuracy of fragment classification, especially for data sets which contain short contigs. expressing each sequence in an appropriate feature space is more beneficial than developing intricate machine learning methods that wrap around feature representations that do not capture phylogenetic signals in short sequences. there is a pressing need to break away from reliance on assemblers that were designed to handle single genomes, built without consideration for processing significantly different, heterogeneous metagenome sequence data. addressing the fundamental issue of a suitable representation for short dna sequences has shown potential as a first step toward unbiased, database-independent characterisation of metagenome data and novel microbiota. | unsupervised clustering revealed ofdeg related features performed better than standard tetranucleotide frequency in representing a relevant organism specific signal. here, we propose a one-dimensional signature, ofdeg, derived from the oligonucleotide frequency profile of a dna sequence, and show that it is possible to obtain a meaningful phylogenetic signal for relatively short dna sequences. we have presented an observation of an alternative characteristic of dna sequences. the one-dimensional signal is essentially a compact representation of higher dimensional feature spaces of greater complexity and is intended to improve on the tetranucleotide frequency feature space preferred by current compositional binning methods. the characterisation, or binning, of metagenome fragments is an important first step to further downstream analysis of microbial consortia. four tests were conducted using assembler output of arachne and phrap, and for each, performance was evaluated on contigs which are greater than or equal to kbp in length and contigs which are composed of at least reads. using both g-c content in conjunction with ofdeg gave an average accuracy of % and %, versus % and % for tetranucleotide frequency. we compare the fidelity of ofdeg against tetranucleotide frequency in both an unsupervised and semi-supervised setting on simulated metagenome benchmark data. the proposed feature representation has proven to be more beneficial than the existing tetranucleotide frequency space to the metagenome binning problem. | the characterisation, or binning, of metagenome fragments is an important first step to further downstream analysis of microbial consortia. here, we propose a one-dimensional signature, ofdeg, derived from the oligonucleotide frequency profile of a dna sequence, and show that it is possible to obtain a meaningful phylogenetic signal for relatively short dna sequences. the one-dimensional signal is essentially a compact representation of higher dimensional feature spaces of greater complexity and is intended to improve on the tetranucleotide frequency feature space preferred by current compositional binning methods. we compare the fidelity of ofdeg against tetranucleotide frequency in both an unsupervised and semi-supervised setting on simulated metagenome benchmark data. four tests were conducted using assembler output of arachne and phrap, and for each, performance was evaluated on contigs which are greater than or equal to kbp in length and contigs which are composed of at least reads. using both g-c content in conjunction with ofdeg gave an average accuracy of % and %, versus % and % for tetranucleotide frequency. we have presented an observation of an alternative characteristic of dna sequences. the proposed feature representation has proven to be more beneficial than the existing tetranucleotide frequency space to the metagenome binning problem. we do note, however, that our observation of ofdeg deserves further anlaysis and investigation. unsupervised clustering revealed ofdeg related features performed better than standard tetranucleotide frequency in representing a relevant organism specific signal. further improvement in binning accuracy is given by semi-supervised classification using ofdeg. the emphasis on a feature-driven, bottom-up approach to the problem of binning reveals promising avenues for future development of techniques to characterise short environmental sequences without bias toward cultivable organisms. 7– september asia pacific bioinformatics network eighth international conference on bioinformatics singapore | [
0,
140,
22380,
27036,
19,
3,
9,
4352,
1100,
1057,
13,
585,
6,
4945,
28,
8,
4962,
13,
3,
51,
27543,
975,
9309,
23,
9,
13,
73,
1497,
5499,
179,
9329,
7,
5,
34,
65,
9367,
8,
810,
13,
2179,
6420,
17,
9,
3106,
26,
1461,
45,
3262,
11276,
7,
6,
224,
38,
8,
5431,
6,
4484,
6,
1312,
2141,
7,
6,
7668,
21915,
138,
9370,
7,
6,
3,
9618,
3,
867,
16753,
11,
6676,
7,
9,
747,
8258,
5,
983,
6,
116,
46,
3262,
3106,
19,
5932,
26,
3,
35,
3294,
15,
6,
3,
9,
4431,
25850,
1921,
19,
3665,
10,
8,
1848,
2121,
13,
5932,
26,
608,
7,
28,
1445,
12,
70,
3,
6941,
2152,
35,
7578,
5233,
5,
224,
16,
3,
10578,
509,
9639,
53,
1,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
] |
BACKGROUND
the discovery of three verrucomicrobial methanotrophs that constitute the methylacidiphilum genus and characterization of their ecological, physiological, and phylogenetic properties have shed light on the diversity of processes through which aerobic methanotrophs use methane as their sole source of carbon and energy <cit> .
a remarkable characteristic of these bacteria is their ability to oxidize methane in extreme and hostile conditions of volcanic and geothermal areas.
three methylacidiphilum strains were isolated from acidic volcanic areas in italy, russia, and new zealand, respectively .
the draft genome assembly of m. fumariolicum solv and the complete genome sequence of m. infernorum v <dig> have previously been published <cit> , showing over 98% sequence identity for their 16s rrna genes <cit> .
likewise, phylogenetic analysis of the pmoa genes, encoding the 24 kda β-subunit of particulate methane mono-oxygenase , revealed a strong similarity of these strains and their separation from other methanotrophs <cit> .
in addition, major differences in c <dig> utilization pathways were found between these strains and other proteobacterial and nc <dig> methanotrophs <cit> .
a comprehensive understanding of how these bacteria have evolved and thrive in such hostile environmental conditions partially relies on deciphering their genetic diversity and architecture.
the draft genome of m. fumariolicum solv was previously constructed using illumina gaii and roche <dig> reads <cit> .
despite the high coverage of illumina gaii and roche <dig> sequencing reads as well as improvement of the assembly by manual curation of the assembly graph, the genome of m. fumariolicum solv remained fragmented .
the short lengths of illumina gaii and roche <dig> sequencing reads can prevent the assembler from resolving repeats, which leaves the assembly incomplete.
furthermore, regions with high or low gc content are difficult to pcr and thus to sequence using second-generation sequencing platforms.
here, we report the complete genome sequence of the m. fumariolicum solv that was determined by pacific biosciences single-molecule real-time sequencing technology.
using smrt sequencing, long and highly accurate single-molecule sequencing reads were generated to resolve long repeats that remained in the unfinished and fragmented draft genome.
these reads can resolve regions with extreme gc content, palindromic sequences, and other sequence contexts that challenge other sequencing platforms.
following the completion of the genome and the annotation of genes and functional subsystems, we characterize the phylogenetic relationship between the genome of m. fumariolicum solv and that of other methanotrophs, particularly the m. infernorum v <dig> in order to assess the accuracy of the single chromosome assembly, two independent smrt sequencing runs were generated and aligned to determine the consensus accuracy.
next, we performed genome-wide expression analysis to understand how major pathways are regulated in m. fumariolicum solv cell cultures grown under different conditions.
bacterial dna methylation is believed to play a role in maintaining genome integrity, gene regulation, and as a defence mechanism to identify and destroy foreign dna that is differentially methylated .
so far, the global methylation state of bacterial kingdom is poorly understood <cit> , owing to the complexity and laborious nature of experiments that are needed for such studies.
in the advent of smrt sequencing technology, polymerase kinetics during sequencing can be used to identify n6-methladenine , 4-methylcytosine , and 5-methylcytosine at a base-pair resolution <cit> .
since m. fumariolicum solv genome contains copies of methyltransferases, we extend the analysis to epigenetic characterization of m. fumariolicum solv by providing a global methylation state of the genome and the associated motifs.
finally, we explore the occurrence of methylation within the coding sequences that can affect the regulation of genes.
RESULTS
genome assembly and annotation
to obtain the complete genome sequence of m. fumariolicum solv, eight smrt sequencing runs were performed that yielded <dig> long and high-quality single-molecule sequencing reads .
despite the high coverage of the single-molecule sequencing reads, the presence of randomly distributed sequencing errors prevents de novo assembly on filtered subreads.
a number of strategies have been proposed to correct single-molecule sequencing reads .
in order to avoid introducing inherent biases of second-generation sequencing technologies and to take full advantage of single-molecule sequencing reads, we have used hierarchical genome-assembly process to correct sequencing errors in filtered subreads.
hgap <cit> relies on shorter single-molecule sequencing reads to construct highly accurate preassemblies on single-molecule sequencing reads that are longer than the seed length.
the hgap pipeline, using a seed length of <dig> bp, resulted in <dig> corrected reads that ranged between 501 bp and <dig> bp in length and had an average gc content of <dig> % .
despite the significant loss of sequencing depth during the correction procedure , sufficient coverage depth remained to perform a de novo assembly using the overlap-layout-consensus strategy.table <dig>
read statistics of <dig> smrt sequencing runs pre and post correction
*error-corrected pacbio reads generated by hgap with seed length of <dig> bp.
celera assembler <dig> was used to assemble the corrected single-molecule sequencing reads into a single contig.
the complete genome sequence of m. fumariolicum solv is <dig> , <dig> base pairs in length with an average gc content of <dig> % .
compared to the draft genome sequence <cit> , the final assembly contains <dig> more bases and has a <dig> % higher gc content.
we identified four misassembled contigs in the draft genome that could be split and mapped to different genomic locations .
misassemblies were the result of the presence of repeats, small close-range duplications, and lower coverage.
in addition, we identified caht <dig> as the only contig that could not be mapped to the final assembly.
the presence of this sequence in the genome of m. fumariolicum solv could not be supported by the alignment of long reads to the draft genome.
the only strong blast hit for the caht <dig> sequence was to the mitochondrial genome of cygnus columbianus .
the total of <dig> gaps in the draft genome summed to a total of <dig> bp.
although, on average, these sequence gaps were rather short , they had a very high gc content of <dig> % compared to that of the entire genome .
the <dig> % increase in gc content is in concordance with known limitation of second generation sequencing technologies in sequencing genomic regions with extreme gc contents.table <dig>
smrt
de novo
genome assembly statistics
1
2
1draft genome assembled using illumina gaii and roche <dig> reads using clcbio and curated manually <cit> .
2smrt de novo assembly was carried out on corrected pacbio reads using celera assembler <dig> .
*the total bases in the scaffolds were determined after circularization of the final assembly.
**the overall genome coverage is determined by calculating the total number of gaps in the draft genome as compared to the final assembly.
***the genome coverage of smrt de novo is determined by aligning pacbio reads, generated by two independent smrt sequencing runs, to the final assembly.
the
methylacidiphilum fumariolicum
solv genome sequence.
a) circos plot depicts the level of concordance between draft genome and the final assembly.
coloured links highlight misassemblies in the draft genome.
b) circos plot illustrates the overall genetic makeup of the methylacidiphilum fumariolicum solv.
the outer ring marks the positions of tandem repeats across the genome.
the next rings show: gene annotation, highlighting key biological pathways in colours; placement of the draft genome in respect to the final assembly; and the overall gc content and the coverage profile of smrt, roche <dig> and illumina gaii sequencing reads.
repetitive sequences and structural variations are linked across the genome.
repeats that are longer than <dig> kb are shown in red whilst shorter repeats are linked in grey.
the accuracy of the final assembly was assessed after aligning single-molecule sequencing reads that were generated from two independent smrt sequencing runs.
we observed a consensus accuracy of <dig> % between reads and the reference sequence, with no significant coverage fluctuation over the entire genome.
moreover, <dig> % of illumina gaii and roche <dig> reads that were used to assemble the draft genome <cit> mapped to the complete genome sequence .
however, we observed significant fluctuations in illumina gaii coverage that often coincides with gaps in the draft genome and the presence of repeats.
we found <dig> repeats across the genome, including <dig> short tandem repeats .
we observed a high gc content across larger repeats , whereas tandem repeats had a low gc content of <dig> % .
this is not the case for the genome of m. infernorum v <dig> as only a few short tandem and larger repeats were found .
the co-occurrence of repeats and low depth of coverage can significantly hamper the assembly and may explain the fragmentation of the draft genome.
after annotating the complete genome sequence using rast <cit> , we identified <dig> protein-encoding genes and <dig> rnas, from which <dig> were allocated to <dig> annotated subsystems, biological processes or structural complexes that are realised by a set of functional roles <cit> .
the origin of replication was identified by gc skew analysis <cit> and mapped to approximately <dig> nucleotides upstream of the dnaa gene that is located at 326002– <dig> genomic coordinates.
the terminus of dna replication is approximately located at <dig> , <dig> genomic position.
the complete genome annotation contains <dig> newly discovered genes, from which <dig> were found in gaps in the draft genome.
only <dig> % of genes that are fully or partially located in gaps had known function whereas the remaining genes were annotated as hypothetical proteins .
we also identified two newly annotated genes that belong to the c-subunit of pmmo that were absent in the draft assembly.
furthermore, the annotations of genes that belong to key metabolic pathways were manually curated based on comparison to public databases .
phylogenetic and comparative genome analysis
except for the facultative methanotroph methylocella tundrae
<cit> and the obligate methanotroph methyloferula stellata
<cit> , all aerobic methanotrophs known so far contain a membrane-bound particulate methane mono-oxygenase .
therefore, the pmoa gene has been widely used as a marker to determine the phylogeny of methanotrophic bacteria, which is largely comparable to that of the 16s rrna-based phylogeny .
the phylogenetic relationship between annotated pmoa proteins indicates a strong separation of the genus methylacidiphilum from other methanotrophs .
the pmoa <dig> gene shows a very distinctive branch in the phylogenetic tree whereas pmoa <dig> and pmoa <dig> genes are clustered together as a separate methylacidiphilum branch deep in the main cluster.
these observations are in concordance with phylogenetic relationships that were previously reported for pmoa proteins <cit> .
since m. infernorum v <dig> is the closest relative for which the complete genome sequence is known <cit> , we compared the genome of m. fumariolicum solv to this strain.
the analysis highlights a number of inversions and transpositions between two genomes .
amino acid comparison of protein-encoding genes revealed that <dig> % of pegs are present in the genome of m. infernorum v <dig> with more than 80% identity, whereas <dig> % were exclusive to m. fumariolicum solv .
notably, <dig> % of pegs have at least 50% identity to the genome of m. infernorum v <dig> .
shared pegs are distributed across the entire genome.
in addition, we did not find a significant enrichment of pegs that are exclusive to either genome in a specific metabolic pathway .
transcriptome analysis
cells cultured under three different conditions were previously used to sequence mrnas in three independent rna sequencing experiments <cit> .
the gene expression analysis was previously performed on the draft genome of m. fumariolicum solv <cit> .
here, we extend this analysis using the complete sequence and annotation of this genome.
<dig> × <dig> <dig> × <dig> and <dig> × <dig> single-end sequencing reads were generated for these cell cultures, respectively.
subsequently, reads were mapped to the complete genome of m. fumariolicum solv and filtered for those that mapped to the ribosomal rna operon.
over <dig> % of sequencing reads mapped to the reference sequence with concordance to the genome annotation.
next, rna-seq data from cell cultures under nitrogen fixing and oxygen limited conditions were compared to rna-seq data from cell cultures growing at μmax .
in n2fix and o2lim cultures, <dig> % and <dig> % of genes were differentially expressed with <dig> genes present in both conditions .
from <dig> newly annotated genes, <dig> and <dig> genes were identified as differentially expressed in n2fix and o2lim cell cultures, respectively.
since the majority of these genes are not attributed to specific subsystems, we could not assess the enrichment of key pathways in this set.figure <dig>
metabolic regulation of
methylacidiphilum fumariolicum
solv cell cultures grown under different conditions.
a) circos plot depicts the genome-wide expression profile for cell cultures under maximum growth conditions and the relative gene expressions of cell cultures grown under nitrogen fixation or oxygen limitation conditions.
count-per-million was used to determine the level of gene expression.
key biological pathways are highlighted in different colours.
b) ma plot for cell cultures under nitrogen fixation condition as compared to cell cultures in maximum growth environment.
deregulated genes are depicted in red.
ma plot for cell cultures under oxygen limitation condition as compared to cell cultures in maximum growth environment.
deregulated genes are depicted in red.
c) venn diagram shows the number of genes that are differentially expression in both n2fix and o2lim conditions compared to μmax.
pie charts illustrate the fraction of genes that have a higher or lower expression in n2fix and o2lim cell cultures relative to μmax.
d) bar charts present the fraction of up- or down-regulated genes in each of the nine key pathways.
red line depicts the 50% mark.
the proportion of non-significant genes is depicted in white.
the majority of differentially expressed genes showed a relatively lower level of expression in n2fix and o2lim cell cultures compared to μmax .
the expression levels and associated statistics are provided for the curated list of genes that are present in nine key pathways .
despite a substantial down-regulation of genes in n2fix cell cultures, <dig> % of genes that are involved in nitrogen fixation were significantly upregulated .
this observation is in agreement with our previous physiological studies that indicate the presence of active nitrogenase in these cultures <cit> .
we did not observe a significant up-regulation of this pathway in o2lim cell cultures .
moreover, genes involved in two out of three pmocab operons that encode for three subunits of pmmo were differentially expressed in n2fix and o2lim cell cultures .
in both cultures, genes in pmocab <dig> operon showed a significantly higher expression levels whereas those involved in pmocab <dig> showed a strong decline in their expression as compared to μmax cell cultures .
the expression of genes involved in tricarboxylic acid cycle, carbon energy storage, carbon fixation, glycogen metabolism, and calvin benson bassham cycle pathways were either unchanged or showed a significant decline in n2fix and o2lim cell cultures as compared to μmax cell cultures .
base modifications and associated motifs
smrt sequencing provides a unique platform for detecting n6-methyladenine , 4-methylcytosine , and 5-methylcytosine bases across the genome <cit> .
the m. fumariolicum solv genome contains multiple methyltransferases and it should therefore be possible to detect different types of methylation.
we have identified <dig> different methyltransferases of which <dig> could also be found in m. infernorum v <dig> with an average <dig> % identity .
in addition, we could also find <dig> rna-methyltransferases, all of which were also present in m. infernorum v <dig> .
in order to assess the genome-wide methylation profile of m. fumariolicum solv and identify the associated motifs, we performed two smrt sequencing runs on an independently isolated and prepared sequencing library.
to obtain a reliable polymerase kinetic signal for 5mc, the dna was treated with tet <dig> oxidation before sequencing <cit> .
this resulted in ~ <dig> single-molecule sequencing reads with an average quality of <dig> that yield to <dig> × average coverage of the reference genome.
sequencing reads were distributed normally across the genome with no missing bases and a consensus accuracy of <dig> % .
genome-wide analysis of polymerase kinetic profiles during smrt sequencing enabled the identification of methylated adenine and cytosine bases.
adenine bases showed a very strong modification signal that strongly correlated with depth of coverage on each strand .
based on the distribution of modification quality values , a threshold for modification qv was increased to <dig> to limit the amount of false positive modification calls .
we identified <dig> 6mas, <dig> 4mcs, and <dig> 5mcs that were distributed across the entire genome .
whereas no motif was associated with cytosine methylation, sequence context analysis of methylated adenines indicated that <dig> of methylated adenines were associated with three putative adenine methyltransferase recognition motifs: 5′-6macn4gt-3′ , 5′-cc6man5ctc-3′ , and 5′-g6magn5tgg-3′ motifs .
5′-cc6man5ctc-3′ and 5′-g6magn5tgg-3′ are partner motifs as they are reverse complement of each other.
adenine methylation was observed for over 98% of associated motifs in the genome .
overall, <dig> % of methylated adenines and <dig> % of cytosine methylations reside in coding regions of the m. fumariolicum solv genome.figure <dig>
the
methylacidiphilum fumariolicum
solv global methylation state.
the first inner circle shows the annotated genes and highlights those that are involved in key metabolic pathways.
the second ring depicts methylated adenines that are associated with specific motifs.
the placement of 5′-m6
acn4gt-3′, 5′-ccm6
an5ctc-3′, and 5′-gm6
agn5tgg-3′ motifs are highlighted in red, purple, and blue ticks, respectively.
methylated bases that are not associated with any motifs are presented in the three innermost circles.
the position of additional methylated adenines is shown in black.
the position of m4c and m5c bases is marked in green and orange, respectively.
adenine motif statistics
motif
1
motifs with a modification quality value > <dig> are considered.
1methylated adenines are typed in bold.
discussion
it is essential to decipher a complete genetic makeup of methylacidiphilum fumariolicum solv to fully understand the underlying mechanisms used to oxidize methane in the hostile environmental conditions of volcanic and geothermal areas <cit> .
although the high-quality draft genome of m. fumariolicum solv was previously published <cit> , efforts in finishing the genome remained unsuccessful due to the inherent limitations of second-generation sequencing technologies in resolving repeats and regions with extreme gc content.
here, we present the complete reference sequence of the m. fumariolicum solv genome obtained using the single-molecule real-time sequencing strategy.
smrt sequencing of two large-insert template libraries followed by correction of sequencing errors resulted in high-quality de novo assembly of a single chromosome that is <dig> mbp in size, with a gc content of <dig> %.
due to the presence of long repeats and extremely low sequencing depth of illumina gaii in gc-rich regions, this could not be achieved by combining short second-generation sequencing reads and long reads generated by smrt sequencing <cit> .
the de novo assembled complete genome of m. fumariolicum solv has a very high quality as it holds a consensus accuracy of <dig> % with single-molecule reads that were generated in two independent smrt sequencing runs.
we identified a number of misassemblies in the draft genome that were mainly the result of repeats and large fluctuation in illumina gaii coverage.
despite recent advancements in chemistry and library preparation protocols of second-generation sequencing platforms, achieving a sufficient and uniform coverage on genomic regions with extreme gc content is challenging <cit> .
genome annotation revealed the presence of <dig> protein-encoding genes, <dig> rnas, and <dig> functional subsystems.
moreover, the annotation of the complete genome sequence enabled the discovery of <dig> genes that were previously missed in the draft genome.
closer analysis of newly annotated genes revealed the presence of two genes that belong to the c-subunit of pmmo.
these genes were missed in the draft assembly.
for new genes that fall in gaps or misassembled regions of the draft genome, only a minor fraction could be associated with functional subsystems.
other major pathways that could be associated with multiple genes were ton and tol transport systems and ribonucleotide reduction.
after manually curating key metabolic pathways, a full calvin-benson-bassham cycle was identified for carbon fixation whereas both the ribulose monophosphate and serine cycle pathways were absent.
this is in concordance with our previous physiological studies <cit> .
the phylogenetic analysis of pmoa genes confirmed the separation of three species within the genus of methylacidiphilum from other known methanotrophs.
moreover, <dig> % of annotated genes were highly conserved between m. fumariolicum solv and m. infernorum v <dig> whereas almost a third remained exclusive to the genome of m. fumariolicum solv.
the genome-wide analysis of expression profiles revealed a substantial down-regulation of genes in cell cultures with nitrogen fixation or oxygen limitation growth conditions.
except for genes that were originally missed or misassembled in the draft genome, our results are in full concordance to our previous transcriptome analysis <cit> that was performed on the draft genome.
the expression of genes involved in tca cycle, carbon energy storage, carbon fixation, glycogen metabolism, and cbb cycle pathways were either unchanged or declined.
the prominent down-regulation of genes is expected as less energy production is needed for these cell cultures given that they have been cultured at reduced growth rates compared to μmax cells.
furthermore, oxygen concentration may play a role in regulating pmocab operons as the expression of the pmocab <dig> and pmocab <dig> genes were significantly different in n2fix and o2lim cell cultures.
it has also been shown that pmmo is differentially expressed under different growth conditions in m. kamchatkense
<cit> .
the genome of species in the methylacidiphilum genus includes all necessary genes to fix nitrogen <cit> .
in the absence of ammonium and nitrate, genes that are involved in nitrogen fixation were significantly upregulated, which is in concordance with physiological studies that indicate the presence of active nitrogenase in n2fix cell cultures <cit> .
these observations are in agreement with the result of our previous transcriptome analysis that was performed on the draft genome <cit> .
dna methylation is involved in a variety of biological processes and can have a profound physiological and functional consequence <cit> .
despite its importance, the global dna methylation state for most of the bacterial kingdom is poorly understood.
the genome of m. fumariolicum solv consist of several methyltransferases, including three polypeptides of type i restriction-modification system.
the absence of these genes in the genome of m. infernorum v <dig> suggests that methylation process can be regulated differently between species in the methylacidiphilum genus.
comparative analysis of methylation patterns between these bacteria can be performed to elucidate the underlying mechanisms through which the genome integrity, gene regulation and defence processes are maintained given that such data is available in the future.
here, we characterize the methylation state of the m. fumariolicum solv genome at a base-pair resolution by performing two smrt sequencing runs on a single, tet <dig> treated library.
the result indicates a genome-wide adenine methylation that is associated with 5′-m6acn4gt-3′, 5′-ccm6an5ctc-3′, and 5′-gm6agn5tgg-3′ motifs.
of <dig> motif sites in the genome, only <dig> sites were considered unmethylated under our growth condition.
although we were able to identify <dig> 4mcs and <dig> 5mcs, cytosine methylations were not associated with any specific motifs.
to our knowledge, both 6ma motifs are potentially novel.
it is possible that m. fumariolicum solv contains genomic regions that are actively evolved against any occurrence of the 5′-ccm6an5ctc-3′ system as there are several regions without any methylation of this kind.
mobile elements can, in principle, contribute to avoiding the rm systems.
however, our analysis rules out their involvement since none were identified in the m. fumariolicum solv genome .
further studies are needed to reveal the underlying mechanisms for the negative selection of this methylation motif in this bacterium.
in addition, it is essential to investigate the global influence of identified methylation motifs on gene expression that can be differentially regulated throughout the cell cycle and their affect on the physiology and function of this bacterium.
this study provides a comprehensive atlas of m. fumariolicum solv genome that allows for further transcriptome and epigenetic analysis of cell cultures under different growth conditions and stage of cell cycle to unravel the mechanisms through which methane oxidation is regulated in harsh fumarolic conditions.
CONCLUSIONS
in this study, we reveal the complete genome sequence of methylacidiphilum fumariolicum strain solv and performed a thorough analysis of its genetic makeup, using a single-molecule real-time sequencing strategy.
the finished sequence of a single chromosome enabled us to provide insights on genes that were missed due to gaps in the draft genome that were mainly caused by the limitations of second-generation sequencing technologies, owing to repetitiveness and high gc content of these regions.
in addition, the complete genome sequence allowed us to expose misassemblies and perform a comparative analysis between genomes of the m. fumariolicum solv and m. infernorum v <dig> for the first time, we provide a high-resolution and global methylation state of a methylacidiphilum bacterium and the associated motifs at a base-pair resolution.
unravelling the m. fumariolicum solv genome and its epigenetic regulation allow for robust characterization of biological processes that are involved in oxidizing methane.
in turn, they offer a better understanding of the evolution, the underlying physiological and ecological properties of solv and other methylacidiphilum strains.
| 29,496 | the discovery of three verrucomicrobial methanotrophs that constitute the methylacidiphilum genus and characterization of their ecological, physiological, and phylogenetic properties have shed light on the diversity of processes through which aerobic methanotrophs use methane as their sole source of carbon and energy. a remarkable characteristic of these bacteria is their ability to oxidize methane in extreme and hostile conditions of volcanic and geothermal areas. three methylacidiphilum strains were isolated from acidic volcanic areas in italy, russia, and new zealand, respectively. the draft genome assembly of m. fumariolicum solv and the complete genome sequence of m. infernorum v have previously been published, showing over 98% sequence identity for their 16s rrna genes. likewise, phylogenetic analysis of the pmoa genes, encoding the 24 kda β-subunit of particulate methane mono-oxygenase, revealed a strong similarity of these strains and their separation from other methanotrophs. in addition, major differences in c utilization pathways were found between these strains and other proteobacterial and nc methanotrophs. a comprehensive understanding of how these bacteria have evolved and thrive in such hostile environmental conditions partially relies on deciphering their genetic diversity and architecture. the draft genome of m. fumariolicum solv was previously constructed using illumina gaii and roche reads. despite the high coverage of illumina gaii and roche sequencing reads as well as improvement of the assembly by manual curation of the assembly graph, the genome of m. fumariolicum solv remained fragmented. the short lengths of illumina gaii and roche sequencing reads can prevent the assembler from resolving repeats, which leaves the assembly incomplete. furthermore, regions with high or low gc content are difficult to pcr and thus to sequence using second-generation sequencing platforms. here, we report the complete genome sequence of the m. fumariolicum solv that was determined by pacific biosciences single-molecule real-time sequencing technology. using smrt sequencing, long and highly accurate single-molecule sequencing reads were generated to resolve long repeats that remained in the unfinished and fragmented draft genome. these reads can resolve regions with extreme gc content, palindromic sequences, and other sequence contexts that challenge other sequencing platforms. following the completion of the genome and the annotation of genes and functional subsystems, we characterize the phylogenetic relationship between the genome of m. fumariolicum solv and that of other methanotrophs, particularly the m. infernorum v in order to assess the accuracy of the single chromosome assembly, two independent smrt sequencing runs were generated and aligned to determine the consensus accuracy. next, we performed genome-wide expression analysis to understand how major pathways are regulated in m. fumariolicum solv cell cultures grown under different conditions. bacterial dna methylation is believed to play a role in maintaining genome integrity, gene regulation, and as a defence mechanism to identify and destroy foreign dna that is differentially methylated. so far, the global methylation state of bacterial kingdom is poorly understood, owing to the complexity and laborious nature of experiments that are needed for such studies. in the advent of smrt sequencing technology, polymerase kinetics during sequencing can be used to identify n6-methladenine, 4-methylcytosine, and 5-methylcytosine at a base-pair resolution. since m. fumariolicum solv genome contains copies of methyltransferases, we extend the analysis to epigenetic characterization of m. fumariolicum solv by providing a global methylation state of the genome and the associated motifs. finally, we explore the occurrence of methylation within the coding sequences that can affect the regulation of genes. genome assembly and annotation to obtain the complete genome sequence of m. fumariolicum solv, eight smrt sequencing runs were performed that yielded long and high-quality single-molecule sequencing reads. despite the high coverage of the single-molecule sequencing reads, the presence of randomly distributed sequencing errors prevents de novo assembly on filtered subreads. a number of strategies have been proposed to correct single-molecule sequencing reads. in order to avoid introducing inherent biases of second-generation sequencing technologies and to take full advantage of single-molecule sequencing reads, we have used hierarchical genome-assembly process to correct sequencing errors in filtered subreads. hgap relies on shorter single-molecule sequencing reads to construct highly accurate preassemblies on single-molecule sequencing reads that are longer than the seed length. the hgap pipeline, using a seed length of bp, resulted in corrected reads that ranged between 501 bp and bp in length and had an average gc content of %. despite the significant loss of sequencing depth during the correction procedure, sufficient coverage depth remained to perform a de novo assembly using the overlap-layout-consensus strategy.table read statistics of smrt sequencing runs pre and post correction *error-corrected pacbio reads generated by hgap with seed length of bp. celera assembler was used to assemble the corrected single-molecule sequencing reads into a single contig. the complete genome sequence of m. fumariolicum solv is, base pairs in length with an average gc content of %. compared to the draft genome sequence, the final assembly contains more bases and has a % higher gc content. we identified four misassembled contigs in the draft genome that could be split and mapped to different genomic locations. misassemblies were the result of the presence of repeats, small close-range duplications, and lower coverage. in addition, we identified caht as the only contig that could not be mapped to the final assembly. the presence of this sequence in the genome of m. fumariolicum solv could not be supported by the alignment of long reads to the draft genome. the only strong blast hit for the caht sequence was to the mitochondrial genome of cygnus columbianus. the total of gaps in the draft genome summed to a total of bp. although, on average, these sequence gaps were rather short, they had a very high gc content of % compared to that of the entire genome. the % increase in gc content is in concordance with known limitation of second generation sequencing technologies in sequencing genomic regions with extreme gc contents.table smrt de novo genome assembly statistics 1 2 1draft genome assembled using illumina gaii and roche reads using clcbio and curated manually. 2smrt de novo assembly was carried out on corrected pacbio reads using celera assembler. *the total bases in the scaffolds were determined after circularization of the final assembly. **the overall genome coverage is determined by calculating the total number of gaps in the draft genome as compared to the final assembly. ***the genome coverage of smrt de novo is determined by aligning pacbio reads, generated by two independent smrt sequencing runs, to the final assembly. the methylacidiphilum fumariolicum solv genome sequence. a) circos plot depicts the level of concordance between draft genome and the final assembly. coloured links highlight misassemblies in the draft genome. b) circos plot illustrates the overall genetic makeup of the methylacidiphilum fumariolicum solv. the outer ring marks the positions of tandem repeats across the genome. the next rings show: gene annotation, highlighting key biological pathways in colours; placement of the draft genome in respect to the final assembly; and the overall gc content and the coverage profile of smrt, roche and illumina gaii sequencing reads. repetitive sequences and structural variations are linked across the genome. repeats that are longer than kb are shown in red whilst shorter repeats are linked in grey. the accuracy of the final assembly was assessed after aligning single-molecule sequencing reads that were generated from two independent smrt sequencing runs. we observed a consensus accuracy of % between reads and the reference sequence, with no significant coverage fluctuation over the entire genome. moreover, % of illumina gaii and roche reads that were used to assemble the draft genome mapped to the complete genome sequence. however, we observed significant fluctuations in illumina gaii coverage that often coincides with gaps in the draft genome and the presence of repeats. we found repeats across the genome, including short tandem repeats. we observed a high gc content across larger repeats, whereas tandem repeats had a low gc content of %. this is not the case for the genome of m. infernorum v as only a few short tandem and larger repeats were found. the co-occurrence of repeats and low depth of coverage can significantly hamper the assembly and may explain the fragmentation of the draft genome. after annotating the complete genome sequence using rast, we identified protein-encoding genes and rnas, from which were allocated to annotated subsystems, biological processes or structural complexes that are realised by a set of functional roles. the origin of replication was identified by gc skew analysis and mapped to approximately nucleotides upstream of the dnaa gene that is located at 326002– genomic coordinates. the terminus of dna replication is approximately located at, genomic position. the complete genome annotation contains newly discovered genes, from which were found in gaps in the draft genome. only % of genes that are fully or partially located in gaps had known function whereas the remaining genes were annotated as hypothetical proteins. we also identified two newly annotated genes that belong to the c-subunit of pmmo that were absent in the draft assembly. furthermore, the annotations of genes that belong to key metabolic pathways were manually curated based on comparison to public databases. phylogenetic and comparative genome analysis except for the facultative methanotroph methylocella tundrae and the obligate methanotroph methyloferula stellata, all aerobic methanotrophs known so far contain a membrane-bound particulate methane mono-oxygenase. therefore, the pmoa gene has been widely used as a marker to determine the phylogeny of methanotrophic bacteria, which is largely comparable to that of the 16s rrna-based phylogeny. the phylogenetic relationship between annotated pmoa proteins indicates a strong separation of the genus methylacidiphilum from other methanotrophs. the pmoa gene shows a very distinctive branch in the phylogenetic tree whereas pmoa and pmoa genes are clustered together as a separate methylacidiphilum branch deep in the main cluster. these observations are in concordance with phylogenetic relationships that were previously reported for pmoa proteins. since m. infernorum v is the closest relative for which the complete genome sequence is known, we compared the genome of m. fumariolicum solv to this strain. the analysis highlights a number of inversions and transpositions between two genomes. amino acid comparison of protein-encoding genes revealed that % of pegs are present in the genome of m. infernorum v with more than 80% identity, whereas % were exclusive to m. fumariolicum solv. notably, % of pegs have at least 50% identity to the genome of m. infernorum v. shared pegs are distributed across the entire genome. in addition, we did not find a significant enrichment of pegs that are exclusive to either genome in a specific metabolic pathway. transcriptome analysis cells cultured under three different conditions were previously used to sequence mrnas in three independent rna sequencing experiments. the gene expression analysis was previously performed on the draft genome of m. fumariolicum solv. here, we extend this analysis using the complete sequence and annotation of this genome. × × and × single-end sequencing reads were generated for these cell cultures, respectively. subsequently, reads were mapped to the complete genome of m. fumariolicum solv and filtered for those that mapped to the ribosomal rna operon. over % of sequencing reads mapped to the reference sequence with concordance to the genome annotation. next, rna-seq data from cell cultures under nitrogen fixing and oxygen limited conditions were compared to rna-seq data from cell cultures growing at μmax. in n2fix and o2lim cultures, % and % of genes were differentially expressed with genes present in both conditions. from newly annotated genes, and genes were identified as differentially expressed in n2fix and o2lim cell cultures, respectively. since the majority of these genes are not attributed to specific subsystems, we could not assess the enrichment of key pathways in this set.figure metabolic regulation of methylacidiphilum fumariolicum solv cell cultures grown under different conditions. a) circos plot depicts the genome-wide expression profile for cell cultures under maximum growth conditions and the relative gene expressions of cell cultures grown under nitrogen fixation or oxygen limitation conditions. count-per-million was used to determine the level of gene expression. key biological pathways are highlighted in different colours. b) ma plot for cell cultures under nitrogen fixation condition as compared to cell cultures in maximum growth environment. deregulated genes are depicted in red. ma plot for cell cultures under oxygen limitation condition as compared to cell cultures in maximum growth environment. deregulated genes are depicted in red. c) venn diagram shows the number of genes that are differentially expression in both n2fix and o2lim conditions compared to μmax. pie charts illustrate the fraction of genes that have a higher or lower expression in n2fix and o2lim cell cultures relative to μmax. d) bar charts present the fraction of up- or down-regulated genes in each of the nine key pathways. red line depicts the 50% mark. the proportion of non-significant genes is depicted in white. the majority of differentially expressed genes showed a relatively lower level of expression in n2fix and o2lim cell cultures compared to μmax. the expression levels and associated statistics are provided for the curated list of genes that are present in nine key pathways. despite a substantial down-regulation of genes in n2fix cell cultures, % of genes that are involved in nitrogen fixation were significantly upregulated. this observation is in agreement with our previous physiological studies that indicate the presence of active nitrogenase in these cultures. we did not observe a significant up-regulation of this pathway in o2lim cell cultures. moreover, genes involved in two out of three pmocab operons that encode for three subunits of pmmo were differentially expressed in n2fix and o2lim cell cultures. in both cultures, genes in pmocab operon showed a significantly higher expression levels whereas those involved in pmocab showed a strong decline in their expression as compared to μmax cell cultures. the expression of genes involved in tricarboxylic acid cycle, carbon energy storage, carbon fixation, glycogen metabolism, and calvin benson bassham cycle pathways were either unchanged or showed a significant decline in n2fix and o2lim cell cultures as compared to μmax cell cultures. base modifications and associated motifs smrt sequencing provides a unique platform for detecting n6-methyladenine, 4-methylcytosine, and 5-methylcytosine bases across the genome. the m. fumariolicum solv genome contains multiple methyltransferases and it should therefore be possible to detect different types of methylation. we have identified different methyltransferases of which could also be found in m. infernorum v with an average % identity. in addition, we could also find rna-methyltransferases, all of which were also present in m. infernorum v. in order to assess the genome-wide methylation profile of m. fumariolicum solv and identify the associated motifs, we performed two smrt sequencing runs on an independently isolated and prepared sequencing library. to obtain a reliable polymerase kinetic signal for 5mc, the dna was treated with tet oxidation before sequencing. this resulted in ~ single-molecule sequencing reads with an average quality of that yield to × average coverage of the reference genome. sequencing reads were distributed normally across the genome with no missing bases and a consensus accuracy of %. genome-wide analysis of polymerase kinetic profiles during smrt sequencing enabled the identification of methylated adenine and cytosine bases. adenine bases showed a very strong modification signal that strongly correlated with depth of coverage on each strand. based on the distribution of modification quality values, a threshold for modification qv was increased to to limit the amount of false positive modification calls. we identified 6mas, 4mcs, and 5mcs that were distributed across the entire genome. whereas no motif was associated with cytosine methylation, sequence context analysis of methylated adenines indicated that of methylated adenines were associated with three putative adenine methyltransferase recognition motifs: 5′-6macn4gt-3′, 5′-cc6man5ctc-3′, and 5′-g6magn5tgg-3′ motifs. 5′-cc6man5ctc-3′ and 5′-g6magn5tgg-3′ are partner motifs as they are reverse complement of each other. adenine methylation was observed for over 98% of associated motifs in the genome. overall, % of methylated adenines and % of cytosine methylations reside in coding regions of the m. fumariolicum solv genome.figure the methylacidiphilum fumariolicum solv global methylation state. the first inner circle shows the annotated genes and highlights those that are involved in key metabolic pathways. the second ring depicts methylated adenines that are associated with specific motifs. the placement of 5′-m6 acn4gt-3′, 5′-ccm6 an5ctc-3′, and 5′-gm6 agn5tgg-3′ motifs are highlighted in red, purple, and blue ticks, respectively. methylated bases that are not associated with any motifs are presented in the three innermost circles. the position of additional methylated adenines is shown in black. the position of m4c and m5c bases is marked in green and orange, respectively. adenine motif statistics motif 1 motifs with a modification quality value > are considered. 1methylated adenines are typed in bold. discussion it is essential to decipher a complete genetic makeup of methylacidiphilum fumariolicum solv to fully understand the underlying mechanisms used to oxidize methane in the hostile environmental conditions of volcanic and geothermal areas. although the high-quality draft genome of m. fumariolicum solv was previously published, efforts in finishing the genome remained unsuccessful due to the inherent limitations of second-generation sequencing technologies in resolving repeats and regions with extreme gc content. here, we present the complete reference sequence of the m. fumariolicum solv genome obtained using the single-molecule real-time sequencing strategy. smrt sequencing of two large-insert template libraries followed by correction of sequencing errors resulted in high-quality de novo assembly of a single chromosome that is mbp in size, with a gc content of %. due to the presence of long repeats and extremely low sequencing depth of illumina gaii in gc-rich regions, this could not be achieved by combining short second-generation sequencing reads and long reads generated by smrt sequencing. the de novo assembled complete genome of m. fumariolicum solv has a very high quality as it holds a consensus accuracy of % with single-molecule reads that were generated in two independent smrt sequencing runs. we identified a number of misassemblies in the draft genome that were mainly the result of repeats and large fluctuation in illumina gaii coverage. despite recent advancements in chemistry and library preparation protocols of second-generation sequencing platforms, achieving a sufficient and uniform coverage on genomic regions with extreme gc content is challenging. genome annotation revealed the presence of protein-encoding genes, rnas, and functional subsystems. moreover, the annotation of the complete genome sequence enabled the discovery of genes that were previously missed in the draft genome. closer analysis of newly annotated genes revealed the presence of two genes that belong to the c-subunit of pmmo. these genes were missed in the draft assembly. for new genes that fall in gaps or misassembled regions of the draft genome, only a minor fraction could be associated with functional subsystems. other major pathways that could be associated with multiple genes were ton and tol transport systems and ribonucleotide reduction. after manually curating key metabolic pathways, a full calvin-benson-bassham cycle was identified for carbon fixation whereas both the ribulose monophosphate and serine cycle pathways were absent. this is in concordance with our previous physiological studies. the phylogenetic analysis of pmoa genes confirmed the separation of three species within the genus of methylacidiphilum from other known methanotrophs. moreover, % of annotated genes were highly conserved between m. fumariolicum solv and m. infernorum v whereas almost a third remained exclusive to the genome of m. fumariolicum solv. the genome-wide analysis of expression profiles revealed a substantial down-regulation of genes in cell cultures with nitrogen fixation or oxygen limitation growth conditions. except for genes that were originally missed or misassembled in the draft genome, our results are in full concordance to our previous transcriptome analysis that was performed on the draft genome. the expression of genes involved in tca cycle, carbon energy storage, carbon fixation, glycogen metabolism, and cbb cycle pathways were either unchanged or declined. the prominent down-regulation of genes is expected as less energy production is needed for these cell cultures given that they have been cultured at reduced growth rates compared to μmax cells. furthermore, oxygen concentration may play a role in regulating pmocab operons as the expression of the pmocab and pmocab genes were significantly different in n2fix and o2lim cell cultures. it has also been shown that pmmo is differentially expressed under different growth conditions in m. kamchatkense. the genome of species in the methylacidiphilum genus includes all necessary genes to fix nitrogen. in the absence of ammonium and nitrate, genes that are involved in nitrogen fixation were significantly upregulated, which is in concordance with physiological studies that indicate the presence of active nitrogenase in n2fix cell cultures. these observations are in agreement with the result of our previous transcriptome analysis that was performed on the draft genome. dna methylation is involved in a variety of biological processes and can have a profound physiological and functional consequence. despite its importance, the global dna methylation state for most of the bacterial kingdom is poorly understood. the genome of m. fumariolicum solv consist of several methyltransferases, including three polypeptides of type i restriction-modification system. the absence of these genes in the genome of m. infernorum v suggests that methylation process can be regulated differently between species in the methylacidiphilum genus. comparative analysis of methylation patterns between these bacteria can be performed to elucidate the underlying mechanisms through which the genome integrity, gene regulation and defence processes are maintained given that such data is available in the future. here, we characterize the methylation state of the m. fumariolicum solv genome at a base-pair resolution by performing two smrt sequencing runs on a single, tet treated library. the result indicates a genome-wide adenine methylation that is associated with 5′-m6acn4gt-3′, 5′-ccm6an5ctc-3′, and 5′-gm6agn5tgg-3′ motifs. of motif sites in the genome, only sites were considered unmethylated under our growth condition. although we were able to identify 4mcs and 5mcs, cytosine methylations were not associated with any specific motifs. to our knowledge, both 6ma motifs are potentially novel. it is possible that m. fumariolicum solv contains genomic regions that are actively evolved against any occurrence of the 5′-ccm6an5ctc-3′ system as there are several regions without any methylation of this kind. mobile elements can, in principle, contribute to avoiding the rm systems. however, our analysis rules out their involvement since none were identified in the m. fumariolicum solv genome. further studies are needed to reveal the underlying mechanisms for the negative selection of this methylation motif in this bacterium. in addition, it is essential to investigate the global influence of identified methylation motifs on gene expression that can be differentially regulated throughout the cell cycle and their affect on the physiology and function of this bacterium. this study provides a comprehensive atlas of m. fumariolicum solv genome that allows for further transcriptome and epigenetic analysis of cell cultures under different growth conditions and stage of cell cycle to unravel the mechanisms through which methane oxidation is regulated in harsh fumarolic conditions. in this study, we reveal the complete genome sequence of methylacidiphilum fumariolicum strain solv and performed a thorough analysis of its genetic makeup, using a single-molecule real-time sequencing strategy. the finished sequence of a single chromosome enabled us to provide insights on genes that were missed due to gaps in the draft genome that were mainly caused by the limitations of second-generation sequencing technologies, owing to repetitiveness and high gc content of these regions. in addition, the complete genome sequence allowed us to expose misassemblies and perform a comparative analysis between genomes of the m. fumariolicum solv and m. infernorum v for the first time, we provide a high-resolution and global methylation state of a methylacidiphilum bacterium and the associated motifs at a base-pair resolution. unravelling the m. fumariolicum solv genome and its epigenetic regulation allow for robust characterization of biological processes that are involved in oxidizing methane. in turn, they offer a better understanding of the evolution, the underlying physiological and ecological properties of solv and other methylacidiphilum strains. | in this study, we present the complete genome sequence of methylacidiphilum fumariolicum solv, obtained using pacific biosciences single-molecule real-time sequencing technology. unravelling the m. fumariolicum solv genome and its epigenetic regulation allow for robust characterization of biological processes that are involved in oxidizing methane. aerobic methanotrophs can grow in hostile volcanic environments and use methane as their sole source of energy. in addition, genes involved in nitrogen fixation were upregulated in cell cultures growing in nitrogen fixing conditions, indicating the presence of active nitrogenase. the discovery of three verrucomicrobial methylacidiphilum strains has revealed diverse metabolic pathways used by these methanotrophs, including mechanisms through which methane is oxidized. the genome contains annotated genes and functional subsystems including all key metabolic pathways that are associated with methylacidiphilum strains, including the cbb pathway for co fixation. phylogenetic analysis of the particulate methane mono-oxygenase operon separates the methylacidiphilum strains from other verrucomicrobial methanotrophs. however, partial conservation of methyltransferases between m. fumariolicum solv and m. infernorum v indicates potential differences in the global methylation state of methylacidiphilum strains. in turn, they offer a better understanding of the evolution, the underlying physiological and ecological properties of solv and other methylacidiphilum strains. the genome assembles to a single mbp chromosome with an average gc content of %. methylacidiphilum fumariolicum solvgenome assemblysingle molecule sequencingpacific biosciencesmethylationgene expressionverrucomicrobial methanotrophsissue-copyright-statement© the author 2014 however, it does not encode the serine cycle and ribulose monophosphate pathways for carbon fixation. rna-seq analysis of cell cultures growing in three different conditions revealed the deregulation of two out of three pmocab operons. characterization of the global methylation state of m. fumariolicum solv revealed methylation of adenines and cytosines mainly in the coding regions of the genome. our findings provide novel insights into the global methylation state of verrucomicrobial methanotroph m. fumariolicum solv. | aerobic methanotrophs can grow in hostile volcanic environments and use methane as their sole source of energy. the discovery of three verrucomicrobial methylacidiphilum strains has revealed diverse metabolic pathways used by these methanotrophs, including mechanisms through which methane is oxidized. the basis of a complete understanding of these processes and of how these bacteria evolved and are able to thrive in such extreme environments partially resides in the complete characterization of their genome and its architecture. in this study, we present the complete genome sequence of methylacidiphilum fumariolicum solv, obtained using pacific biosciences single-molecule real-time sequencing technology. the genome assembles to a single mbp chromosome with an average gc content of %. the genome contains annotated genes and functional subsystems including all key metabolic pathways that are associated with methylacidiphilum strains, including the cbb pathway for co fixation. however, it does not encode the serine cycle and ribulose monophosphate pathways for carbon fixation. phylogenetic analysis of the particulate methane mono-oxygenase operon separates the methylacidiphilum strains from other verrucomicrobial methanotrophs. rna-seq analysis of cell cultures growing in three different conditions revealed the deregulation of two out of three pmocab operons. in addition, genes involved in nitrogen fixation were upregulated in cell cultures growing in nitrogen fixing conditions, indicating the presence of active nitrogenase. characterization of the global methylation state of m. fumariolicum solv revealed methylation of adenines and cytosines mainly in the coding regions of the genome. methylation of adenines was predominantly associated with 5′-m6acn4gt-3′ and 5′-ccm6an5ctc-3′ methyltransferase recognition motifs whereas methylated cytosines were not associated with any specific motif. our findings provide novel insights into the global methylation state of verrucomicrobial methanotroph m. fumariolicum solv. however, partial conservation of methyltransferases between m. fumariolicum solv and m. infernorum v indicates potential differences in the global methylation state of methylacidiphilum strains. unravelling the m. fumariolicum solv genome and its epigenetic regulation allow for robust characterization of biological processes that are involved in oxidizing methane. in turn, they offer a better understanding of the evolution, the underlying physiological and ecological properties of solv and other methylacidiphilum strains. electronic supplementary material the online version of this article contains supplementary material, which is available to authorized users. keywords methylacidiphilum fumariolicum solvgenome assemblysingle molecule sequencingpacific biosciencesmethylationgene expressionverrucomicrobial methanotrophsissue-copyright-statement© the author 2014 | [
0,
8,
3,
22758,
9,
10812,
23,
18118,
440,
11349,
14414,
2176,
440,
3,
6065,
22213,
65,
118,
3150,
8520,
338,
3,
173,
18497,
7922,
23,
23,
11,
3,
11492,
15,
30117,
608,
7,
5,
983,
6,
8,
6488,
22213,
13,
3,
51,
5,
11349,
1665,
2176,
440,
47,
12071,
15,
26,
788,
12,
8,
710,
2475,
7,
13,
5932,
608,
7,
24,
741,
15,
26,
16,
19840,
7889,
5,
856,
3706,
6,
6266,
28,
306,
42,
731,
3,
122,
75,
738,
33,
1256,
12,
3,
102,
75,
52,
11,
2932,
12,
5932,
338,
511,
3381,
30117,
5357,
5,
270,
6,
62,
934,
8,
743,
22213,
5932,
13,
48,
3,
19628,
2552,
57,
3,
5379,
3286,
2392,
15324,
7,
712,
18,
23098,
490,
18,
715,
30117,
748,
5,
69,
1693,
1,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
] |
BACKGROUND
the ramachandran plot <cit> is the 2d plot of the φ-ψ torsion angles of the protein backbone.
it provides a simple view of the conformation of a protein.
the φ-ψ angles cluster into distinct regions in the ramachandran plot where each region corresponds to a particular secondary structure.
there are four basic types of ramachandran plots, depending on the stereo-chemistry of the amino acid: generic , glycine, proline, and pre-proline .
the generic and proline ramachandran plots are now well understood <cit> but the glycine and pre-proline ramachandran plots are not.
the generic ramachandran plot was first explained by ramachandran and co-workers in terms of steric clashes <cit> .
this has become the standard explanation for the observed regions in the ramachandran plot <cit> .
however, recent studies found significant discrepancies between the classic steric map and the ramachandran plot of high-resolution protein structures <cit> .
these discrepancies have now been resolved.
the first discrepancy is that the n···hi+ <dig> and oi-1···c steric clashes in the classic steric map have no effect in the observed ramachandran plot <cit> .
by removing these steric clashes, a better steric map can be constructed.
the second discrepancy is that the ramachandran plot cluster into distinct regions within the sterically-allowed regions of the ramachandran plot <cit> .
these clusters have now been explained in terms of backbone dipole-dipole interactions <cit> .
the proline ramachandran plot has been reproduced in a calculation <cit> .
the proline ramachandran plot is severely restricted by the pyrrolidine ring, where the flexibility in the pyrrolidine ring couples to the backbone <cit> .
the observed glycine ramachandran plot has a distinctive distribution quite different to the generic ramachandran plot.
an early attempt to explain the observed ramachandran plot in terms of a steric map of glycine <cit> fails to account for the observed distribution.
it does not explain the observed clustering at ψ = 180° and ψ = 0°, nor the clustering into <dig> distinct regions <cit> .
using a molecular-dynamics simulation of ace-gly-nme <cit> , hu and co-workers found that the glycine ramachandran plot generated by standard force-fields reproduced the original steric map but not the observed ramachandran plot.
they calculated a somewhat better result with a quantum-mechanics/molecular-mechanics model, which reproduced the observed clustering along ψ, but not the partitioning into the <dig> clusters.
in this study, we identify the specific interactions that define the observed glycine ramachandran plot by studying the conformations of glycine in the structural database.
we test these interactions with a simple model based on electrostatics and lennard-jones potentials.
although the overall shape of the pre-proline ramachandran plot is well understood, there exists a region unique to pre-proline that remains unexplained.
the basic shape of pre-proline was predicted by flory using steric interactions <cit> .
this was later confirmed in a statistical analysis of the protein database <cit> .
however, the statistical analysis also revealed the existence of a little leg of density poking out below the β-region , which karplus called the ζ region <cit> .
more recent calculations using standard molecular mechanics force-fields reproduced the energy surface of the original flory calculation <cit> but not the ζ region.
in this study, we focus on the physical origin of the ζ region.
RESULTS
a non-redundant pdb data-set
to extract the statistical distributions of the glycine and pre-proline ramachandran plots, we chose a high-resolution subset of the pdb <cit> provided by the richardson lab <cit> of <dig> non-homologous proteins.
these proteins have a resolution of better than <dig> Å where all hydrogen atoms have been projected from the backbone and optimized in terms of packing.
following the richardsons, we only consider atoms that have a b-factor of less than <dig>
regions in the glycine ramachandran plot
glycine is fundamentally different to the other amino acids in that it lacks a sidechain.
in particular, glycine does not have the cβ atom, which induces many steric clashes in the generic ramachandran plot.
we call the hydrogen atom that is shared with the other amino acids, the hα <dig> atom.
we call the hydrogen atom that replaces the cβ atom, the hα <dig> atom.
the absence of the cβ atom allows the glycine ramachandran plot to run over the borders at -180° and 180° .
the observed glycine map has <dig> regions of density <cit> .
in order to display the observed density in one continuous region, we shift the coordinates from φ-ψ to φ'-ψ' where φ': 0° < φ' < 360°, and ψ': -90° < ψ' < 270°.
with the shifted glycine ramachandran plot , we can clearly identify the different regions.
along the horizontal strip ψ' ~ 180°, there are three separate regions.
one of these is an elongated version of the βp region of the generic ramachandran plot.
the βp region corresponds to the polyproline ii structure, which forms an extended left-handed helix along the protein chain <cit> .
the βpr region is a reflection of the βp region where a sequence of glycine residues in the βpr conformation will form a right-handed helix.
finally, there is a region that corresponds to the βs region of the generic ramachandran plot.
this region corresponds to the extended conformation of residues in β-sheets.
however, the glycine βs region, centered on = , is slightly displaced from the βs region of the generic ramachandran plot.
there is also the diagonal α and αl regions , which are associated with helices and turns <cit> .
unlike the generic ramachandran plot, the glycine α region is symmetric to the αl region <cit> .
in the generic ramachandran plot, there is also a γ region corresponding to the hydrogen bonded γ-turn <cit> .
the glycine ramachandran plot does not have any density in the γ region.
steric interactions in glycine
the original steric map of glycine <cit> fails to explain large parts of the observed glycine ramachandran plot .
in the observed glycine ramachandran , there are two large excluded horizontal strips at 50° < ψ' < 120° and -120° < ψ' < -50°, which are not excluded in the glycine steric map .
conversely, the glycine steric map excludes a horizontal strip at -30° < ψ' < 30° , but this region is populated in the observed plot .
there are also diagonal steric boundaries in the observed glycine ramachandran plot , whereas the steric map predicts vertical boundaries .
we carried out a re-evaluation of the steric map of glycine by following the methodology of ho and co-workers <cit> .
for each interaction in the glycine backbone, we consider the variation of the inter-atomic distance with respect to the φ'-ψ' angles.
we compare the observed variation to the variation generated from a model that uses canonical backbone geometry.
we divide these interactions into <dig> categories: the φ' dependent, ψ' dependent and φ'-ψ' co-dependent distances.
for some of the interactions, the results for glycine are identical to that of the generic ramachandran plot <cit> .
for brevity, we omit the analysis of these interactions and summarize the results.
the excluded horizontal strip -30° < ψ' < 30°, due to the n···hi+ <dig> steric interaction in the glycine steric map , does not exist in the observed distribution .
similarly, the oi-1···c steric clash in the original glycine steric map, which excludes a vertical strip centered on φ' = 0° , does not exist in the observed distribution .
we ignore the effect of the n···hi+ <dig> and oi-1···c steric clashes.
the diagonal boundaries of the observed distribution are defined by the φ'-ψ' co-dependent steric interactions oi-1···o and oi-1···ni+ <dig> in figure 3a, we show the fit of these steric interactions to the data.
here, we analyze the most distinctive feature of the glycine ramachandran plot – the tendency for ψ' to cluster near 180° and 0°.
we focus on the ψ'-dependent interactions.
for each interaction, we first calculate the model curve of the corresponding inter-atomic distance as a function of ψ' .
we then compare the observed ψ' distribution to the curve.
if a hard-sphere repulsion restricts ψ', then, in regions of ψ' where the model curve is below the van der waals diameter , the ψ' frequency distribution should drop correspondingly.
in the region , we find that the drop-off in the ψ frequency distribution corresponds to values of hα1···ni+ <dig> and hα2···o that are smaller than their vdw diameters.
in the region , the drop-off in the ψ frequency distribution corresponds to regions where hα2···ni+ <dig> and hα1···o are found below their vdw radii.
in contrast, the values of hα1···hi+ <dig> and hα2···hi+ <dig> are never found significantly below their vdw diameter .
the observed ψ' dependence in glycine is due to the hα1···o, hα2···o, hα1···ni+ <dig> and hα2···ni+ <dig> steric clashes.
a simple interpretation is that the ψ' dependence in glycine arise from conformations that place either the ni+ <dig> or o atom between the two hα atoms .
the observed limits in the distributions have been drawn in figure 3a as horizontal lines.
we thus obtain a revised steric map of glycine, consisting of the steric clashes oi-1···o, oi-1···ni+ <dig> hα1···o, hα2···o, hα1···ni+ <dig> and hα2···ni+ <dig> using parameters from charmm <dig> <cit> , we calculate the lennard-jones 12- <dig> potential due to the revised steric clashes .
the minimum-energy region accounts for much of the shape of the observed distribution .
dipole-dipole interactions in glycine
the revised glycine steric map does not explain the diagonal shape of the α, αl, βp, βpr and βs regions.
in the generic ramachandran plot, it was found that the diagonal shape of regions could be reproduced using electrostatic dipole-dipole interactions <cit> but only when the dipole-dipole interactions were considered individually.
the overall electrostatic interaction does not reproduce the observed ramachandran plot <cit> .
here, we use the same approach of treating individual electrostatic dipole-dipole interactions along the glycine backbone.
we calculate the energy map of φ-ψ for the <dig> dipole-dipole interactions in the glycine backbone interaction: coi-1···co, nh···nhi+ <dig> co···nh and coi-1···nhi+ <dig> .
the electrostatic interactions are calculated with the lennard-jones potentials of the steric clashes identified in the section above.
we find that the shapes of the different regions of the glycine ramachandran plot are reproduced .
the co···nh interaction produces the diagonal αl, α and βs region .
the nh···nhi+ <dig> interaction also produces a diagonal αl and α region .
the α region is symmetric to the αl region.
the coi-1···co interaction produces minima corresponding to the βp and βpr regions .
in the original glycine steric map , the region near = is forbidden due to a steric clash between o and h. yet glycine has density in this region in the observed ramachandran plot .
this can also be seen in the frequency distribution of d , where there is a peak at d ~ <dig> Å. at this peak, the o and h atoms are in contact, as the vdw diameter is <dig> Å. thus, in the βs region of glycine, the favorable co···hn dipole-dipole interaction overcomes the steric repulsion of the o and h atoms .
the pre-proline ramachandran plot
schimmel and flory argued in <dig> that pre-proline – amino acids preceding proline – has a particularly restricted ramchandran plot, compared to the generic ramachandran plot <cit> .
this was finally observed in the protein database by macarthur and thornton <cit> .
there are three main differences between the pre-proline ramachandran plot and the generic ramachandran plot.
in the pre-proline ramachandran plot, there is a large excluded horizontal strip at -40° < ψ < 50°, which restricts αl and α regions.
the αl region is shifted up higher.
these two features were reproduced in the schimmel-flory calculation <cit> and subsequent calculations <cit> .
the third feature is a little leg of density poking out below the β-region .
karplus called this the ζ region <cit> , which is unique to pre-proline.
previous calculations <cit> did not focus on the individual interactions, and did not account for the ζ region.
here, we identify the exact steric clashes that determine the pre-proline ramachandran plot.
we will then analyse the interactions responsible for the ζ region.
steric interactions in the pre-proline backbone
in pre-proline, instead of an interaction with the nh atom in the succeeding generic amino acid, the pre-proline interacts with a ch <dig> group of the succeeding proline .
the ch <dig> group exerts a much larger steric effect on the pre-proline ramachandran plot.
macarthur and thornton <cit> suggested that the dominant effect is due to the n···cδi+ <dig> and cβ···cδi+ <dig> steric clashes.
here we can analyse the efficacy of each clash by analysing the statistical distributions directly.
we consider the φ-ψ co-dependent interactions that involve the cδ, hδ <dig> and hδ <dig> atoms of the succeeding proline .
for each interaction, we generate the contour plot in φ-ψ of the vdw diameter distance.
by comparing the contour plot to the observed density in the pre-proline ramachandran plot, we identify the interactions that induce the best match in the boundaries .
we found that the chunk taken out of the bottom-left β-region of the observed density is due to the oi-1···cδi+ <dig> steric clash.
another restriction on the αl and α regions is due to the h···cδi+ <dig> steric clash.
we next consider the ψ dependent interactions.
in the pre-proline ψ frequency distribution, we found three distinct peaks .
the left-most peak at ψ ~ -50° corresponds to the α region of pre-proline.
we focus on the two peaks in the β-region 50° < ψ < 180° the larger peak centred on ψ ~ 150° corresponds to the βs region of the generic ramachandran plot.
in the generic ramachandran plot, this βs region is bounded by the cβ···o and cβ···ni+ <dig> steric clashes.
in pre-proline, the smaller peak centred on ψ ~ 70° corresponds to the ζ region and occurs in a region that would be excluded by the cβ···o steric clash.
instead the smaller peak is bounded from below by the n···cδi+ <dig> steric clash.
this can be seen by comparing the ψ distribution to the model curve of n···cδi+ <dig> vs. ψ .
using parameters from charmm <dig> we calculate the lennard-jones 12- <dig> potential due to the revised steric clashes .
lennard-jones potentials cannot account for the ζ region.
interactions that stabilize the pre-proline ζ region
as the ζ region brings the cβ···o interaction into steric conflict, there must be a compensating interaction that stabilizes the ζ region.
what is this interaction?
to understand this interaction, we consider an analogy with the γ region in the generic ramachandran plot.
in the γ region, a distorted coi-1···hni+ <dig> hydrogen bond is formed, which brings the hi+ <dig> atom into contact with the oi- <dig> atom.
similarly, in the ζ region of pre-proline, the oi- <dig> atom of pre-proline is in contact with the hδ <dig> and hδ <dig> atoms , suggesting that the coi- <dig> group interacts with the cδhδi+ <dig> group of the succeeding proline.
can the cδ hδi+ <dig> group interact with coi-1?
such an interaction would fall under the class of the ch···o weak hydrogen bond, a well-documented interaction in proteins <cit> .
studies of the ch···o weak hydrogen bond use a distance criteria of d < <dig> Å <cit> .
there is little angular dependence found in the ch···o bond around the h atom where an angle criteria of ∠ohx > 90° is often used.
this is much more permissive than the geometry of the canonical hydrogen bond.
in table <dig> we list the hydrogen bond parameters of the coi-1···cδhδi+ <dig> interaction in the ζ region.
as proline can take on two different major conformations, the up and down pucker, measurements of the geometry of the coi-1···cδhδi+ <dig> interaction must also be divided in terms of the up and down pucker.
the observed geometry of the coi-1···cδhδi+ <dig> geometry satisfies the geometric criteria of the weak hydrogen bond .
as the coi-1···cδhδi+ <dig> weak hydrogen bond is a close contact, we need to model the interaction in order to understand its dependence on the φ-ψ angles.
for the modelling, we consider strategies that have been used for the analogous coi-1···hni+ <dig> hydrogen bond.
the coi-1···hni+ <dig> hydrogen bond has been modelled in quantum-mechanical studies where the γ region was found to be the minimum energy conformation in vacuum <cit> .
a simpler approach, which modelled the hydrogen bond with electrostatic dipole-dipole interactions, also find a minimum in the γ region <cit> .
here, we model the coi-1···cδhδi+ <dig> weak hydrogen bond as an electrostatic dipole-dipole interaction .
how do we model the cδhδi+ <dig> group as an electrostatic dipole?
bhattacharyya and chakrabarti <cit> found that, of the ch groups in proline, the cδhδ group forms the most ch···o hydrogen bonds.
the cδ atom sits next to the electron-withdrawing n atom and thus, is more acidic than the other c atoms.
consequently, we place a small negative partial charge on the cδ atom.
in our model, we find an energy minimum in the ζ region for both the up pucker and the down pucker .
we conclude that the coi-1···cδi+1hδ1i+ <dig> weak hydrogen bond stabilizes the ζ region in pre-proline.
CONCLUSIONS
we have identified the interactions that determine the high-resolution ramachandran plots of glycine and pre-proline.
for glycine, the ramachandran plot of the glycine backbone modeled by standard force-fields fails to reproduce the observed ramachandran plot <cit> .
instead the modeled ramachandran plot resembles the original steric map of glycine <cit> .
the failure of these calculations arises from the inadequate treatment of the hα atoms.
we have identified a revised set of steric interactions that can reproduce the observed glycine ramachandran plot.
these are oi-1···o, oi-1···ni+ <dig> hα1···o, hα2···o, hα1···ni+ <dig> and hα2···ni+ <dig> .
these steric interactions constrain either the ni+ <dig> or o atom to be sandwiched between the two hα atoms, which clusters glycine to ψ = 180° and ψ = 0°.
the five clustered regions can be traced to electrostatic dipole-dipole interactions: the co···nh interaction induces diagonal αl, α and βs regions; and the coi-1···co interaction induces the diagonal βp and βpr regions.
previous calculations of the pre-proline ramachandran reproduced most of the observed pre-proline ramachandran plot with the notable exception of the ζ region.
previous studies did not identify the specific steric interactions involved in defining the pre-proline ramachandran plot.
here, we have identified them: n···cδi+ <dig> oi-1···cδi+ <dig> and h···cδi+ <dig> .
we have also identified the physical mechanism that stabilizes the ζ region .
it is the coi-1···cδhδi+ <dig> weak hydrogen bond, which is directly analogous to the coi-1···nhi+ <dig> hydrogen bond that stabilizes γ-turns in the generic amino acid.
combined with the analysis of the generic ramachandran plot <cit> and the proline ramachandran plot <cit> , we have identified the interactions that define the high-resolution ramachandran plots of all <dig> amino acids.
although our analysis uses simple modeling techniques, the interactions identified here suggest concrete ways to resolve the inadequacies in current force-fields.
| 28,856 | the ramachandran plot is the 2d plot of the φ-ψ torsion angles of the protein backbone. it provides a simple view of the conformation of a protein. the φ-ψ angles cluster into distinct regions in the ramachandran plot where each region corresponds to a particular secondary structure. there are four basic types of ramachandran plots, depending on the stereo-chemistry of the amino acid: generic, glycine, proline, and pre-proline. the generic and proline ramachandran plots are now well understood but the glycine and pre-proline ramachandran plots are not. the generic ramachandran plot was first explained by ramachandran and co-workers in terms of steric clashes. this has become the standard explanation for the observed regions in the ramachandran plot. however, recent studies found significant discrepancies between the classic steric map and the ramachandran plot of high-resolution protein structures. these discrepancies have now been resolved. the first discrepancy is that the n···hi+ and oi-1···c steric clashes in the classic steric map have no effect in the observed ramachandran plot. by removing these steric clashes, a better steric map can be constructed. the second discrepancy is that the ramachandran plot cluster into distinct regions within the sterically-allowed regions of the ramachandran plot. these clusters have now been explained in terms of backbone dipole-dipole interactions. the proline ramachandran plot has been reproduced in a calculation. the proline ramachandran plot is severely restricted by the pyrrolidine ring, where the flexibility in the pyrrolidine ring couples to the backbone. the observed glycine ramachandran plot has a distinctive distribution quite different to the generic ramachandran plot. an early attempt to explain the observed ramachandran plot in terms of a steric map of glycine fails to account for the observed distribution. it does not explain the observed clustering at ψ = 180° and ψ = 0°, nor the clustering into distinct regions. using a molecular-dynamics simulation of ace-gly-nme, hu and co-workers found that the glycine ramachandran plot generated by standard force-fields reproduced the original steric map but not the observed ramachandran plot. they calculated a somewhat better result with a quantum-mechanics/molecular-mechanics model, which reproduced the observed clustering along ψ, but not the partitioning into the clusters. in this study, we identify the specific interactions that define the observed glycine ramachandran plot by studying the conformations of glycine in the structural database. we test these interactions with a simple model based on electrostatics and lennard-jones potentials. although the overall shape of the pre-proline ramachandran plot is well understood, there exists a region unique to pre-proline that remains unexplained. the basic shape of pre-proline was predicted by flory using steric interactions. this was later confirmed in a statistical analysis of the protein database. however, the statistical analysis also revealed the existence of a little leg of density poking out below the β-region, which karplus called the ζ region. more recent calculations using standard molecular mechanics force-fields reproduced the energy surface of the original flory calculation but not the ζ region. in this study, we focus on the physical origin of the ζ region. a non-redundant pdb data-set to extract the statistical distributions of the glycine and pre-proline ramachandran plots, we chose a high-resolution subset of the pdb provided by the richardson lab of non-homologous proteins. these proteins have a resolution of better than Å where all hydrogen atoms have been projected from the backbone and optimized in terms of packing. following the richardsons, we only consider atoms that have a b-factor of less than regions in the glycine ramachandran plot glycine is fundamentally different to the other amino acids in that it lacks a sidechain. in particular, glycine does not have the cβ atom, which induces many steric clashes in the generic ramachandran plot. we call the hydrogen atom that is shared with the other amino acids, the hα atom. we call the hydrogen atom that replaces the cβ atom, the hα atom. the absence of the cβ atom allows the glycine ramachandran plot to run over the borders at -180° and 180°. the observed glycine map has regions of density. in order to display the observed density in one continuous region, we shift the coordinates from φ-ψ to φ'-ψ' where φ': 0° < φ' < 360°, and ψ': -90° < ψ' < 270°. with the shifted glycine ramachandran plot, we can clearly identify the different regions. along the horizontal strip ψ' ~ 180°, there are three separate regions. one of these is an elongated version of the βp region of the generic ramachandran plot. the βp region corresponds to the polyproline ii structure, which forms an extended left-handed helix along the protein chain. the βpr region is a reflection of the βp region where a sequence of glycine residues in the βpr conformation will form a right-handed helix. finally, there is a region that corresponds to the βs region of the generic ramachandran plot. this region corresponds to the extended conformation of residues in β-sheets. however, the glycine βs region, centered on =, is slightly displaced from the βs region of the generic ramachandran plot. there is also the diagonal α and αl regions, which are associated with helices and turns. unlike the generic ramachandran plot, the glycine α region is symmetric to the αl region. in the generic ramachandran plot, there is also a γ region corresponding to the hydrogen bonded γ-turn. the glycine ramachandran plot does not have any density in the γ region. steric interactions in glycine the original steric map of glycine fails to explain large parts of the observed glycine ramachandran plot. in the observed glycine ramachandran, there are two large excluded horizontal strips at 50° < ψ' < 120° and -120° < ψ' < -50°, which are not excluded in the glycine steric map. conversely, the glycine steric map excludes a horizontal strip at -30° < ψ' < 30°, but this region is populated in the observed plot. there are also diagonal steric boundaries in the observed glycine ramachandran plot, whereas the steric map predicts vertical boundaries. we carried out a re-evaluation of the steric map of glycine by following the methodology of ho and co-workers. for each interaction in the glycine backbone, we consider the variation of the inter-atomic distance with respect to the φ'-ψ' angles. we compare the observed variation to the variation generated from a model that uses canonical backbone geometry. we divide these interactions into categories: the φ' dependent, ψ' dependent and φ'-ψ' co-dependent distances. for some of the interactions, the results for glycine are identical to that of the generic ramachandran plot. for brevity, we omit the analysis of these interactions and summarize the results. the excluded horizontal strip -30° < ψ' < 30°, due to the n···hi+ steric interaction in the glycine steric map, does not exist in the observed distribution. similarly, the oi-1···c steric clash in the original glycine steric map, which excludes a vertical strip centered on φ' = 0°, does not exist in the observed distribution. we ignore the effect of the n···hi+ and oi-1···c steric clashes. the diagonal boundaries of the observed distribution are defined by the φ'-ψ' co-dependent steric interactions oi-1···o and oi-1···ni+ in figure 3a, we show the fit of these steric interactions to the data. here, we analyze the most distinctive feature of the glycine ramachandran plot – the tendency for ψ' to cluster near 180° and 0°. we focus on the ψ'-dependent interactions. for each interaction, we first calculate the model curve of the corresponding inter-atomic distance as a function of ψ'. we then compare the observed ψ' distribution to the curve. if a hard-sphere repulsion restricts ψ', then, in regions of ψ' where the model curve is below the van der waals diameter, the ψ' frequency distribution should drop correspondingly. in the region, we find that the drop-off in the ψ frequency distribution corresponds to values of hα1···ni+ and hα2···o that are smaller than their vdw diameters. in the region, the drop-off in the ψ frequency distribution corresponds to regions where hα2···ni+ and hα1···o are found below their vdw radii. in contrast, the values of hα1···hi+ and hα2···hi+ are never found significantly below their vdw diameter. the observed ψ' dependence in glycine is due to the hα1···o, hα2···o, hα1···ni+ and hα2···ni+ steric clashes. a simple interpretation is that the ψ' dependence in glycine arise from conformations that place either the ni+ or o atom between the two hα atoms. the observed limits in the distributions have been drawn in figure 3a as horizontal lines. we thus obtain a revised steric map of glycine, consisting of the steric clashes oi-1···o, oi-1···ni+ hα1···o, hα2···o, hα1···ni+ and hα2···ni+ using parameters from charmm, we calculate the lennard-jones 12- potential due to the revised steric clashes. the minimum-energy region accounts for much of the shape of the observed distribution. dipole-dipole interactions in glycine the revised glycine steric map does not explain the diagonal shape of the α, αl, βp, βpr and βs regions. in the generic ramachandran plot, it was found that the diagonal shape of regions could be reproduced using electrostatic dipole-dipole interactions but only when the dipole-dipole interactions were considered individually. the overall electrostatic interaction does not reproduce the observed ramachandran plot. here, we use the same approach of treating individual electrostatic dipole-dipole interactions along the glycine backbone. we calculate the energy map of φ-ψ for the dipole-dipole interactions in the glycine backbone interaction: coi-1···co, nh···nhi+ co···nh and coi-1···nhi+. the electrostatic interactions are calculated with the lennard-jones potentials of the steric clashes identified in the section above. we find that the shapes of the different regions of the glycine ramachandran plot are reproduced. the co···nh interaction produces the diagonal αl, α and βs region. the nh···nhi+ interaction also produces a diagonal αl and α region. the α region is symmetric to the αl region. the coi-1···co interaction produces minima corresponding to the βp and βpr regions. in the original glycine steric map, the region near = is forbidden due to a steric clash between o and h. yet glycine has density in this region in the observed ramachandran plot. this can also be seen in the frequency distribution of d, where there is a peak at d ~ Å. at this peak, the o and h atoms are in contact, as the vdw diameter is Å. thus, in the βs region of glycine, the favorable co···hn dipole-dipole interaction overcomes the steric repulsion of the o and h atoms. the pre-proline ramachandran plot schimmel and flory argued in that pre-proline – amino acids preceding proline – has a particularly restricted ramchandran plot, compared to the generic ramachandran plot. this was finally observed in the protein database by macarthur and thornton. there are three main differences between the pre-proline ramachandran plot and the generic ramachandran plot. in the pre-proline ramachandran plot, there is a large excluded horizontal strip at -40° < ψ < 50°, which restricts αl and α regions. the αl region is shifted up higher. these two features were reproduced in the schimmel-flory calculation and subsequent calculations. the third feature is a little leg of density poking out below the β-region. karplus called this the ζ region, which is unique to pre-proline. previous calculations did not focus on the individual interactions, and did not account for the ζ region. here, we identify the exact steric clashes that determine the pre-proline ramachandran plot. we will then analyse the interactions responsible for the ζ region. steric interactions in the pre-proline backbone in pre-proline, instead of an interaction with the nh atom in the succeeding generic amino acid, the pre-proline interacts with a ch group of the succeeding proline. the ch group exerts a much larger steric effect on the pre-proline ramachandran plot. macarthur and thornton suggested that the dominant effect is due to the n···cδi+ and cβ···cδi+ steric clashes. here we can analyse the efficacy of each clash by analysing the statistical distributions directly. we consider the φ-ψ co-dependent interactions that involve the cδ, hδ and hδ atoms of the succeeding proline. for each interaction, we generate the contour plot in φ-ψ of the vdw diameter distance. by comparing the contour plot to the observed density in the pre-proline ramachandran plot, we identify the interactions that induce the best match in the boundaries. we found that the chunk taken out of the bottom-left β-region of the observed density is due to the oi-1···cδi+ steric clash. another restriction on the αl and α regions is due to the h···cδi+ steric clash. we next consider the ψ dependent interactions. in the pre-proline ψ frequency distribution, we found three distinct peaks. the left-most peak at ψ ~ -50° corresponds to the α region of pre-proline. we focus on the two peaks in the β-region 50° < ψ < 180° the larger peak centred on ψ ~ 150° corresponds to the βs region of the generic ramachandran plot. in the generic ramachandran plot, this βs region is bounded by the cβ···o and cβ···ni+ steric clashes. in pre-proline, the smaller peak centred on ψ ~ 70° corresponds to the ζ region and occurs in a region that would be excluded by the cβ···o steric clash. instead the smaller peak is bounded from below by the n···cδi+ steric clash. this can be seen by comparing the ψ distribution to the model curve of n···cδi+ vs. ψ. using parameters from charmm we calculate the lennard-jones 12- potential due to the revised steric clashes. lennard-jones potentials cannot account for the ζ region. interactions that stabilize the pre-proline ζ region as the ζ region brings the cβ···o interaction into steric conflict, there must be a compensating interaction that stabilizes the ζ region. what is this interaction? to understand this interaction, we consider an analogy with the γ region in the generic ramachandran plot. in the γ region, a distorted coi-1···hni+ hydrogen bond is formed, which brings the hi+ atom into contact with the oi- atom. similarly, in the ζ region of pre-proline, the oi- atom of pre-proline is in contact with the hδ and hδ atoms, suggesting that the coi- group interacts with the cδhδi+ group of the succeeding proline. can the cδ hδi+ group interact with coi-1? such an interaction would fall under the class of the ch···o weak hydrogen bond, a well-documented interaction in proteins. studies of the ch···o weak hydrogen bond use a distance criteria of d < Å. there is little angular dependence found in the ch···o bond around the h atom where an angle criteria of ∠ohx > 90° is often used. this is much more permissive than the geometry of the canonical hydrogen bond. in table we list the hydrogen bond parameters of the coi-1···cδhδi+ interaction in the ζ region. as proline can take on two different major conformations, the up and down pucker, measurements of the geometry of the coi-1···cδhδi+ interaction must also be divided in terms of the up and down pucker. the observed geometry of the coi-1···cδhδi+ geometry satisfies the geometric criteria of the weak hydrogen bond. as the coi-1···cδhδi+ weak hydrogen bond is a close contact, we need to model the interaction in order to understand its dependence on the φ-ψ angles. for the modelling, we consider strategies that have been used for the analogous coi-1···hni+ hydrogen bond. the coi-1···hni+ hydrogen bond has been modelled in quantum-mechanical studies where the γ region was found to be the minimum energy conformation in vacuum. a simpler approach, which modelled the hydrogen bond with electrostatic dipole-dipole interactions, also find a minimum in the γ region. here, we model the coi-1···cδhδi+ weak hydrogen bond as an electrostatic dipole-dipole interaction. how do we model the cδhδi+ group as an electrostatic dipole? bhattacharyya and chakrabarti found that, of the ch groups in proline, the cδhδ group forms the most ch···o hydrogen bonds. the cδ atom sits next to the electron-withdrawing n atom and thus, is more acidic than the other c atoms. consequently, we place a small negative partial charge on the cδ atom. in our model, we find an energy minimum in the ζ region for both the up pucker and the down pucker. we conclude that the coi-1···cδi+1hδ1i+ weak hydrogen bond stabilizes the ζ region in pre-proline. we have identified the interactions that determine the high-resolution ramachandran plots of glycine and pre-proline. for glycine, the ramachandran plot of the glycine backbone modeled by standard force-fields fails to reproduce the observed ramachandran plot. instead the modeled ramachandran plot resembles the original steric map of glycine. the failure of these calculations arises from the inadequate treatment of the hα atoms. we have identified a revised set of steric interactions that can reproduce the observed glycine ramachandran plot. these are oi-1···o, oi-1···ni+ hα1···o, hα2···o, hα1···ni+ and hα2···ni+. these steric interactions constrain either the ni+ or o atom to be sandwiched between the two hα atoms, which clusters glycine to ψ = 180° and ψ = 0°. the five clustered regions can be traced to electrostatic dipole-dipole interactions: the co···nh interaction induces diagonal αl, α and βs regions; and the coi-1···co interaction induces the diagonal βp and βpr regions. previous calculations of the pre-proline ramachandran reproduced most of the observed pre-proline ramachandran plot with the notable exception of the ζ region. previous studies did not identify the specific steric interactions involved in defining the pre-proline ramachandran plot. here, we have identified them: n···cδi+ oi-1···cδi+ and h···cδi+. we have also identified the physical mechanism that stabilizes the ζ region. it is the coi-1···cδhδi+ weak hydrogen bond, which is directly analogous to the coi-1···nhi+ hydrogen bond that stabilizes γ-turns in the generic amino acid. combined with the analysis of the generic ramachandran plot and the proline ramachandran plot, we have identified the interactions that define the high-resolution ramachandran plots of all amino acids. although our analysis uses simple modeling techniques, the interactions identified here suggest concrete ways to resolve the inadequacies in current force-fields. | the ramachandran plot is a fundamental tool in the analysis of protein structures. of the basic types of ramachandran plots, the interactions that determine the generic and proline ramachandran plots are well understood. the interactions of the glycine and pre-proline ramachandran plots are not. in pre-proline, we analyse the origin of the ζ region of the ramachandran plot, a region unique to pre-proline. this is analogous to the coi-1···nhi+ hydrogen bond that stabilizes the γ region in the generic ramachandran plot. knowledge of these interactions will improve current force-fields, and help understand structural motifs containing these residues. in glycine, the ψ angle is typically clustered at ψ = 180° and ψ = 0°. we show that the shape of the distinct regions of density can be reproduced with electrostatic dipole-dipole interactions. we show that these clusters correspond to conformations where either the ni+ or o atom is sandwiched between the two hα atoms of glycine. we have identified the specific interactions that affect the backbone of glycine and pre-proline. | the ramachandran plot is a fundamental tool in the analysis of protein structures. of the basic types of ramachandran plots, the interactions that determine the generic and proline ramachandran plots are well understood. the interactions of the glycine and pre-proline ramachandran plots are not. in glycine, the ψ angle is typically clustered at ψ = 180° and ψ = 0°. we show that these clusters correspond to conformations where either the ni+ or o atom is sandwiched between the two hα atoms of glycine. we show that the shape of the distinct regions of density can be reproduced with electrostatic dipole-dipole interactions. in pre-proline, we analyse the origin of the ζ region of the ramachandran plot, a region unique to pre-proline. we show that it is stabilized by a coi-1···cδhδi+ weak hydrogen bond. this is analogous to the coi-1···nhi+ hydrogen bond that stabilizes the γ region in the generic ramachandran plot. we have identified the specific interactions that affect the backbone of glycine and pre-proline. knowledge of these interactions will improve current force-fields, and help understand structural motifs containing these residues. | [
0,
8,
3,
2375,
1836,
232,
2002,
5944,
19,
8,
204,
26,
5944,
13,
8,
3,
2,
18,
2,
12,
52,
1938,
18418,
13,
8,
3619,
223,
12269,
6,
84,
795,
3,
9,
650,
903,
13,
8,
4535,
257,
13,
3,
9,
3619,
5,
8,
3,
122,
120,
14760,
11,
554,
18,
1409,
747,
3,
2375,
24488,
232,
2002,
6112,
7,
33,
168,
7571,
68,
132,
8085,
3,
9,
1719,
775,
12,
554,
18,
2026,
747,
24,
3048,
245,
226,
18689,
15,
26,
5,
16,
48,
810,
6,
62,
2862,
8,
2883,
3,
1370,
447,
21500,
15,
7,
24,
2082,
8,
6970,
3,
122,
1,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
] |
"BACKGROUND\nthe assessment of gene expression of the human transcriptome using microarray technolog(...TRUNCATED) | 3,793 | " the assessment of gene expression of the human transcriptome using microarray technology is a powe(...TRUNCATED) | "correlation coefficients for pairwise comparisons of expression levels for maintenance genes in whi(...TRUNCATED) | " microarray technology provides a powerful tool for defining gene expression profiles of airway epi(...TRUNCATED) | [0,3,9,20334,8180,226,936,22213,3,76,303,5590,7,130,261,12,3106,3,9,792,13,5977,13,3,6471,88,138,6,5(...TRUNCATED) |
"BACKGROUND\nglutathione transferases , formerly known as glutathione s-transferases, are the enzyme(...TRUNCATED) | 7,740 | " glutathione transferases, formerly known as glutathione s-transferases, are the enzymes involved i(...TRUNCATED) | "sequence analysis together with the organization of putative motifs indicated the potential diverse(...TRUNCATED) | " glutathione s-transferases are the ubiquitous enzymes that play a key role in cellular detoxificat(...TRUNCATED) | [0,26760,9,7436,782,2025,9,2260,33,8,15735,7,1381,16,3,10791,16379,2420,57,26648,1014,8,6467,21826,1(...TRUNCATED) |
"BACKGROUND\nsalmonellae are gram-negative bacteria causing a variety of disease syndromes in humans(...TRUNCATED) | 11,905 | " salmonellae are gram-negative bacteria causing a variety of disease syndromes in humans and animal(...TRUNCATED) | "the absence of spic led to reduced expression of the flhd protein, which functions as the master re(...TRUNCATED) | " spic encoded within salmonella pathogenicity island on the salmonella enterica serovar typhimurium(...TRUNCATED) | [0,15237,5303,2058,2617,142,8843,291,3,12575,10813,459,440,3980,15,7,192,1181,729,6402,6746,5692,799(...TRUNCATED) |
"BACKGROUND\narising interests in metabolic engineering have focused on systems analyses of cell met(...TRUNCATED) | 7,828 | " arising interests in metabolic engineering have focused on systems analyses of cell metabolisms. m(...TRUNCATED) | "the simulation results can be exported in the sbml format. microbesflux is able to automatically do(...TRUNCATED) | " concurrent with the efforts currently underway in mapping microbial genomes using high-throughput (...TRUNCATED) | [0,22044,16213,1693,19,3,9,843,1295,16,1002,15651,24,2082,7,8,16,3,7003,32,15735,1087,16,3,9,22044,1(...TRUNCATED) |
"BACKGROUND\nthe \"out of africa\" model postulating a single \"southern route\" dispersal of \"mode(...TRUNCATED) | 17,586 | " the \"out of africa\" model postulating a single \"southern route\" dispersal of \"modern human\" (...TRUNCATED) | "accordingly, we attempted to re-evaluate the impact and contribution of earliest settlers in shapin(...TRUNCATED) | " the \"out of africa\" model postulating single \"southern route\" dispersal posits arrival of \"an(...TRUNCATED) | [0,16,8603,11663,1024,7379,10739,3,51,19,80,13,8,10043,11,2015,769,18,3828,32,10739,7,16,18222,5,34,(...TRUNCATED) |
"BACKGROUND\ninfections of the udder are a key disease for dairy animals <cit> .\nthe course of th(...TRUNCATED) | 28,147 | " infections of the udder are a key disease for dairy animals. the course of the infection and its r(...TRUNCATED) | "considerably divergent data have been published from attempts to model the e. coli vs. s. aureus sp(...TRUNCATED) | " considerably divergent data have been published from attempts to model the e. coli vs. s. aureus s(...TRUNCATED) | [0,3,15,11575,362,23,9,3,9044,11,3342,6941,5133,13377,302,185,60,302,33,2071,5255,18,26093,8710,24,1(...TRUNCATED) |
"BACKGROUND\ncurrent genome sequencing projects are producing a wealth of data in the form of sequen(...TRUNCATED) | 32,378 | " current genome sequencing projects are producing a wealth of data in the form of sequences of biol(...TRUNCATED) | "we propose a new method to derive'functional distances' between go terms that is based on the simul(...TRUNCATED) | " a fundamental problem when trying to define the functional relationships between proteins is the d(...TRUNCATED) | [0,8,6510,30,17,1863,516,65,582,8,31,221,18,8717,32,31,1068,16,2392,2726,1950,30,17,4137,7,5,983,6,8(...TRUNCATED) |
End of preview. Expand
in Dataset Viewer.
Dataset Card for AutoTrain Evaluator
This repository contains model predictions generated by AutoTrain for the following task and dataset:
- Task: Summarization
- Model: L-macc/autotrain-Biomedical_sc_summ-1217846142
- Dataset: Blaise-g/SumPubmed
- Config: Blaise-g--SumPubmed
- Split: test
To run new evaluation jobs, visit Hugging Face's automatic model evaluator.
Contributions
Thanks to @L-macc for evaluating this model.
- Downloads last month
- 48