add initial curation code
Browse files- .gitignore +1 -0
- src/00_setup_curation.sh +29 -0
- src/01_gather_data.R +19 -0
- src/02.1_assemble_K50_dG_dataset.R +142 -0
- src/02.8_gather_ThermoMPNN_splits.py +20 -0
.gitignore
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
*~
|
src/00_setup_curation.sh
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
|
3 |
+
|
4 |
+
|
5 |
+
# from a base directory
|
6 |
+
|
7 |
+
|
8 |
+
mkdir data
|
9 |
+
mkdir intermediate
|
10 |
+
|
11 |
+
git clone https://<user_name>:<security_token>huggingface.co/RosettaCommons/MegaScale
|
12 |
+
|
13 |
+
# needed to get splits
|
14 |
+
git clone https://github.com/Kuhlman-Lab/ThermoMPNN.git
|
15 |
+
|
16 |
+
|
17 |
+
|
18 |
+
|
19 |
+
# Run each numbered script in MegaScale/src/ in order (starting with this one)
|
20 |
+
#
|
21 |
+
# Tips:
|
22 |
+
# 1) Make sure to set the working directory to the base directory (outside of the HF repo)
|
23 |
+
# 2) While most of the scripts should work, I recommend running them interactively
|
24 |
+
# 3) Some stages require more memory than others, all can be done with < 400GB of memory
|
25 |
+
# but perhaps more more could reduce memory requirements
|
26 |
+
# 4)
|
27 |
+
|
28 |
+
|
29 |
+
|
src/01_gather_data.R
ADDED
@@ -0,0 +1,19 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
|
3 |
+
# Download data from: https://zenodo.org/records/7992926
|
4 |
+
# Mega-scale experimental analysis of protein folding stability in biology and design
|
5 |
+
|
6 |
+
# 1.2 GB
|
7 |
+
system("cd data && curl -o data.zip https://zenodo.org/api/records/7992926/files-archive")
|
8 |
+
|
9 |
+
md5sum_expected <- "b76b523d10e12d34b481916b1f57f31c data/data.zip"
|
10 |
+
md5sum <- system(
|
11 |
+
"md5sum data/data.zip",
|
12 |
+
intern = TRUE)
|
13 |
+
if (md5sum != md5sum_expected) {
|
14 |
+
cat("Expected and obtained md5sum values don't match\n")
|
15 |
+
}
|
16 |
+
|
17 |
+
system("cd data && unzip data.zip && rm data.zip")
|
18 |
+
|
19 |
+
|
src/02.1_assemble_K50_dG_dataset.R
ADDED
@@ -0,0 +1,142 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
system("cd data; unzip Processed_K50_dG_datasets.zip")
|
2 |
+
|
3 |
+
|
4 |
+
### Dataset1 ###
|
5 |
+
dataset1 <- readr::read_csv(
|
6 |
+
file = "data/Processed_K50_dG_datasets/Tsuboyama2023_Dataset1_20230416.csv",
|
7 |
+
col_types = readr::cols(
|
8 |
+
name = readr::col_character(),
|
9 |
+
dna_seq = readr::col_character(),
|
10 |
+
log10_K50_t = readr::col_double(),
|
11 |
+
log10_K50_t_95CI_high = readr::col_double(),
|
12 |
+
log10_K50_t_95CI_low = readr::col_double(),
|
13 |
+
log10_K50_t_95CI = readr::col_double(),
|
14 |
+
fitting_error_t = readr::col_double(),
|
15 |
+
log10_K50unfolded_t = readr::col_double(),
|
16 |
+
deltaG_t = readr::col_double(),
|
17 |
+
deltaG_t_95CI_high = readr::col_double(),
|
18 |
+
deltaG_t_95CI_low = readr::col_double(),
|
19 |
+
deltaG_t_95CI = readr::col_double(),
|
20 |
+
log10_K50_c = readr::col_double(),
|
21 |
+
log10_K50_c_95CI_high = readr::col_double(),
|
22 |
+
log10_K50_c_95CI_low = readr::col_double(),
|
23 |
+
log10_K50_c_95CI = readr::col_double(),
|
24 |
+
fitting_error_c = readr::col_double(),
|
25 |
+
log10_K50unfolded_c = readr::col_double(),
|
26 |
+
deltaG_c = readr::col_double(),
|
27 |
+
deltaG_c_95CI_high = readr::col_double(),
|
28 |
+
deltaG_c_95CI_low = readr::col_double(),
|
29 |
+
deltaG_c_95CI = readr::col_double(),
|
30 |
+
deltaG = readr::col_double(),
|
31 |
+
deltaG_95CI_high = readr::col_double(),
|
32 |
+
deltaG_95CI_low = readr::col_double(),
|
33 |
+
deltaG_95CI = readr::col_double(),
|
34 |
+
log10_K50_trypsin_ML = readr::col_double(),
|
35 |
+
log10_K50_chymotrypsin_ML = readr::col_double()))
|
36 |
+
|
37 |
+
# note that some of the log10_K50_trypsin_ML and log10_K50_chmotrypsin_ML values are "-" and ">2.5".
|
38 |
+
# These are parsed as NA values"
|
39 |
+
|
40 |
+
dataset1 |>
|
41 |
+
arrow::write_parquet(
|
42 |
+
"intermediate/dataset1.parquet")
|
43 |
+
|
44 |
+
|
45 |
+
|
46 |
+
dataset23 <- readr::read_csv(
|
47 |
+
file = "data/Processed_K50_dG_datasets/Tsuboyama2023_Dataset2_Dataset3_20230416.csv"
|
48 |
+
|
49 |
+
|
50 |
+
####
|
51 |
+
|
52 |
+
system("cd data && unzip AlphaFold_model_PDBs.zip")
|
53 |
+
|
54 |
+
|
55 |
+
assemble_models <- function(
|
56 |
+
data_path,
|
57 |
+
dataset_tag,
|
58 |
+
pattern,
|
59 |
+
output_path) {
|
60 |
+
|
61 |
+
cat(
|
62 |
+
"data path: ", data_path, "\n",
|
63 |
+
"dataset_tag: ", dataset_tag, "\n",
|
64 |
+
"pattern: ", pattern, "\n",
|
65 |
+
"output path: ", output_path, "\n",
|
66 |
+
sep = "")
|
67 |
+
|
68 |
+
file_index <- 1
|
69 |
+
models <- list.files(
|
70 |
+
path = data_path,
|
71 |
+
full.names = TRUE,
|
72 |
+
pattern = pattern,
|
73 |
+
recursive = TRUE) |>
|
74 |
+
purrr::map_dfr(.f = function(path) {
|
75 |
+
file_handle <- path |>
|
76 |
+
file(open = "rb") |>
|
77 |
+
gzcon()
|
78 |
+
|
79 |
+
if( file_index %% 10 == 0) {
|
80 |
+
cat("Reading '", path, "' ", file_index, "\n", sep = "")
|
81 |
+
}
|
82 |
+
file_index <<- file_index + 1
|
83 |
+
|
84 |
+
lines <- file_handle |> readLines()
|
85 |
+
file_handle |> close()
|
86 |
+
|
87 |
+
data.frame(
|
88 |
+
dataset_tag = dataset_tag,
|
89 |
+
id = path |> basename() |> stringr::str_replace(".pdb", ""),
|
90 |
+
pdb = lines |> paste0(collapse = "\n"))
|
91 |
+
})
|
92 |
+
models |> arrow::write_parquet(output_path)
|
93 |
+
}
|
94 |
+
|
95 |
+
|
96 |
+
assemble_models(
|
97 |
+
data_path = "data/AlphaFold_model_PDBs",
|
98 |
+
dataset_tag = "all",
|
99 |
+
pattern = "*.pdb",
|
100 |
+
output_path = "intermediate/all_pdbs.parquet")
|
101 |
+
|
102 |
+
#
|
103 |
+
# assemble_models(
|
104 |
+
# data_path = "data/AlphaFold_model_PDBs",
|
105 |
+
# dataset_tag = "EA",
|
106 |
+
# pattern = "EA[:]run.*pdb",
|
107 |
+
# output_path = "intermediate/EA_pdbs.parquet")
|
108 |
+
#
|
109 |
+
#
|
110 |
+
# assemble_models(
|
111 |
+
# data_path = "data/AlphaFold_model_PDBs",
|
112 |
+
# dataset_tag = "EEHEE",
|
113 |
+
# pattern = "EEHEE.*pdb",
|
114 |
+
# output_path = "intermediate/EEHEE_pdbs.parquet")
|
115 |
+
#
|
116 |
+
#
|
117 |
+
# assemble_models(
|
118 |
+
# data_path = "data/AlphaFold_model_PDBs",
|
119 |
+
# dataset_tag = "EHEE",
|
120 |
+
# pattern = "EHEE.*pdb",
|
121 |
+
# output_path = "intermediate/EHEE_pdbs.parquet")
|
122 |
+
#
|
123 |
+
#
|
124 |
+
# assemble_models(
|
125 |
+
# data_path = "data/AlphaFold_model_PDBs",
|
126 |
+
# dataset_tag = "GG",
|
127 |
+
# pattern = "GG[:]run.*pdb",
|
128 |
+
# output_path = "intermediate/GG_pdbs.parquet")
|
129 |
+
#
|
130 |
+
#
|
131 |
+
# assemble_models(
|
132 |
+
# data_path = "data/AlphaFold_model_PDBs",
|
133 |
+
# dataset_tag = "HEEH_KT",
|
134 |
+
# pattern = "HEEH_KT_rd.*pdb",
|
135 |
+
# output_path = "intermediate/HEEH_KT_pdbs.parquet")
|
136 |
+
#
|
137 |
+
# assemble_models(
|
138 |
+
# data_path = "data/AlphaFold_model_PDBs",
|
139 |
+
# dataset_tag = "HEEH",
|
140 |
+
# pattern = "HEEH_rd.*pdb",
|
141 |
+
# output_path = "intermediate/HEEH_pdbs.parquet")
|
142 |
+
|
src/02.8_gather_ThermoMPNN_splits.py
ADDED
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
|
3 |
+
import pandas
|
4 |
+
import pyarrow.parquet
|
5 |
+
import pickle
|
6 |
+
|
7 |
+
with open("ThermoMPNN/dataset_splits/mega_splits.pkl", "rb") as f:
|
8 |
+
mega_splits = pickle.load(f)
|
9 |
+
|
10 |
+
splits = []
|
11 |
+
for split_name, split_ids in mega_splits.items():
|
12 |
+
splits.append(
|
13 |
+
pandas.DataFrame({
|
14 |
+
'split_name': split_name,
|
15 |
+
'id': split_ids}))
|
16 |
+
|
17 |
+
splits = pandas.concat(splits)
|
18 |
+
pyarrow.parquet.write_table(
|
19 |
+
pyarrow.Table.from_pandas(splits),
|
20 |
+
where = "intermediate/ThermoMPNN_splits.parquet")
|