Dataset Card for neuralbioinfo/PhaStyle-SequenceDB

Dataset Summary

The PhaStyle-SequenceDB dataset consists of phage sequences labeled with their corresponding lifestyles (virulent or temperate). The dataset is split into four key subsets:

• BACPHLIP training set: 1868 sequences for model training
• BACPHLIP validation set: 394 Escherichia coli sequences for validation
• EXTREMOPHILE set: 16 sequences from extreme environments
• ESCHERICHIA set: Guelin collection + 100 randomly selected high-quality temperate phages

This dataset is specifically designed to aid in training and evaluating genomic models, such as ProkBERT, for phage lifestyle prediction.

Dataset Structure

Figure 1: The dataset used in the ProkBERT PhaStyle study. Phage sequences from multiple independent data sources were segmented into 512bp and 1022bp fragments for training and testing models on phage lifestyle prediction. The dataset consists of the BACPHLIP training and validation sets, Escherichia phages (from the Guelin collection), and phages from extreme environments.

The structure of the dataset is explained visually in the following figure:

• BACPHLIP Training Set: This subset consists of 1,868 phage sequences collected from non-Escherichia phages. The sequences are carefully labeled as virulent or temperate, and the dataset was constructed using the high-quality annotations from the BACPHLIP study. These sequences are used to train the ProkBERT model for lifestyle prediction.

• BACPHLIP Validation Set: Comprising 246 Escherichia phage sequences, this validation set is used to test the performance of the model after training. The sequences are also labeled based on their lifestyle and are critical for evaluating the generalization capability of the ProkBERT model on Escherichia species.

• ESCHERICHIA (Guelin collection + 100 high-quality phages): This dataset contains 394 Escherichia phages, which includes the Guelin collection and an additional 100 randomly selected, high-quality temperate phages. The sequences are segmented into smaller chunks (512bp and 1022bp), representing real-world metagenome assemblies, often fragmented.

• EXTREMOPHILE Phage Set: This test dataset consists of 16 bacteriophages isolated from extreme environments like deep-sea, acidic environments, and arsenic-rich microbial mats. It represents underexplored environments, making it essential for testing how well the models generalize to novel phages outside the commonly studied species.

Data Fields

• sequence_id: Unique identifier for each sequence.
• dataset: The specific dataset the sequence belongs to (e.g., ESCHERICHIA, EXTREMOPHILE, BACPHLIP_TRAINING).
• description: Additional details about the sequence, such as its classification or length.
• label: Indicates the lifestyle of the phage (virulent or temperate).
• y: Numerical encoding of the lifestyle (1 for virulent, 0 for temperate).
• sequence: The nucleotide sequence of the phage.

Dataset Creation

The sequences in this dataset were gathered from various sources, including the BACPHLIP database and curated collections of phages from extreme environments. Each sequence was carefully segmented into smaller fragments (512bp or 1022bp) to simulate real-world scenarios where phage sequences are often fragmented. The training data excludes Escherichia sequences, which are used in the test set to evaluate model generalization capabilities.

Intended Uses

This dataset is intended for use in phage lifestyle prediction tasks using genomic language models such as ProkBERT. The segmented sequences allow models to generalize well even with fragmented or out-of-sample data. It is particularly useful for applications in ecological and clinical settings where understanding phage behavior is critical.

Citing this work

If you use the data in this package, please cite:

@Article{ProkBERT2024,
  author  = {Ligeti, Balázs and Szepesi-Nagy, István and Bodnár, Babett and Ligeti-Nagy, Noémi and Juhász, János},
  journal = {Frontiers in Microbiology},
  title   = {{ProkBERT} family: genomic language models for microbiome applications},
  year    = {2024},
  volume  = {14},
  URL={https://www.frontiersin.org/articles/10.3389/fmicb.2023.1331233},       
    DOI={10.3389/fmicb.2023.1331233},      
    ISSN={1664-302X}
}