Upload process_atomica.py
Browse files- process_atomica.py +240 -0
process_atomica.py
ADDED
@@ -0,0 +1,240 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import tempfile
|
3 |
+
import logging
|
4 |
+
import requests
|
5 |
+
import pandas as pd
|
6 |
+
from tqdm import tqdm
|
7 |
+
from rdkit import Chem
|
8 |
+
import pooch
|
9 |
+
from concurrent.futures import ThreadPoolExecutor, as_completed
|
10 |
+
from lobster.data import upload_to_s3
|
11 |
+
|
12 |
+
logging.basicConfig(level=logging.INFO)
|
13 |
+
logger = logging.getLogger(__name__)
|
14 |
+
|
15 |
+
S3_BASE_URI = os.environ["S3_BASE_URI"]
|
16 |
+
S3_RAW = f"{S3_BASE_URI}/raw"
|
17 |
+
S3_PROCESSED = f"{S3_BASE_URI}/pre-processed"
|
18 |
+
|
19 |
+
MAX_WORKERS = 16 # for threading
|
20 |
+
|
21 |
+
MODALITY_MAPPING = {
|
22 |
+
"polypeptide(L)": "amino_acid",
|
23 |
+
"polynucleotide": "nucleotide",
|
24 |
+
"polyribonucleotide": "nucleotide",
|
25 |
+
"polynucleotide (RNA)": "nucleotide",
|
26 |
+
"polynucleotide (DNA)": "nucleotide",
|
27 |
+
"polydeoxyribonucleotide": "nucleotide",
|
28 |
+
"polydeoxyribonucleotide/polyribonucleotide hybrid ": "nucleotide",
|
29 |
+
"smiles": "smiles",
|
30 |
+
}
|
31 |
+
|
32 |
+
IDENTIFIERS = {
|
33 |
+
"protein-peptide": "11033993",
|
34 |
+
"protein-rna": "11033983",
|
35 |
+
"protein-protein": "11033984",
|
36 |
+
"rna-small_molecule": "11033985",
|
37 |
+
"protein-small_molecule": "11033996",
|
38 |
+
"protein-ion": "11033989",
|
39 |
+
"protein-dna": "11033982",
|
40 |
+
"small_molecule-small_molecule": "11033997",
|
41 |
+
}
|
42 |
+
|
43 |
+
def upload_raw_datasets_to_s3() -> None:
|
44 |
+
for fname, identifier in IDENTIFIERS.items():
|
45 |
+
with tempfile.TemporaryDirectory() as tmpdir:
|
46 |
+
local_path = pooch.retrieve(
|
47 |
+
f"https://dataverse.harvard.edu/api/access/datafile/{identifier}",
|
48 |
+
fname=f"{fname}.csv",
|
49 |
+
known_hash=None,
|
50 |
+
path=tmpdir,
|
51 |
+
progressbar=True,
|
52 |
+
)
|
53 |
+
upload_to_s3(
|
54 |
+
s3_uri=f"{S3_RAW}/{fname}.csv",
|
55 |
+
local_filepath=local_path,
|
56 |
+
)
|
57 |
+
|
58 |
+
|
59 |
+
def _canonicalize_smiles(smiles: str) -> str | None:
|
60 |
+
"""Canonicalize SMILES string."""
|
61 |
+
mol = Chem.MolFromSmiles(smiles)
|
62 |
+
return Chem.MolToSmiles(mol, canonical=True) if mol else None
|
63 |
+
|
64 |
+
|
65 |
+
def _fetch_pdb_data(pdb_id: str) -> list[dict]:
|
66 |
+
"""Fetch molecular data from PDBe API for a given PDB ID."""
|
67 |
+
res = requests.get(f"https://www.ebi.ac.uk/pdbe/api/pdb/entry/molecules/{pdb_id}")
|
68 |
+
|
69 |
+
if res.status_code != 200:
|
70 |
+
raise Exception(f"Failed to fetch data for {pdb_id}: {res.status_code}")
|
71 |
+
|
72 |
+
return res.json().get(pdb_id.lower(), [])
|
73 |
+
|
74 |
+
|
75 |
+
def _fetch_smiles_data(ligand_id: str) -> str | None:
|
76 |
+
"""Fetch SMILES string for ligand from PDBe."""
|
77 |
+
try:
|
78 |
+
res = requests.get(f"https://www.ebi.ac.uk/pdbe/api/pdb/compound/summary/{ligand_id.lower()}")
|
79 |
+
|
80 |
+
if res.status_code != 200:
|
81 |
+
raise Exception(f"Failed to fetch data for {ligand_id}: {res.status_code}")
|
82 |
+
|
83 |
+
ligand_data = res.json().get(ligand_id)[0]
|
84 |
+
ligand_data = iter(ligand_data["smiles"])
|
85 |
+
while True:
|
86 |
+
try:
|
87 |
+
smiles = next(ligand_data)["name"]
|
88 |
+
return _canonicalize_smiles(smiles)
|
89 |
+
except StopIteration:
|
90 |
+
break
|
91 |
+
except Exception:
|
92 |
+
continue
|
93 |
+
except Exception as e:
|
94 |
+
raise Exception(f"Failed to fetch SMILES for {ligand_id}") from e
|
95 |
+
|
96 |
+
|
97 |
+
def _safe_extract(row_id: str, sequence_extractor: callable) -> dict | None:
|
98 |
+
"""Safely extract sequences and modalities given a row ID and extractor."""
|
99 |
+
try:
|
100 |
+
sequences, modalities = sequence_extractor(row_id)
|
101 |
+
except Exception as e:
|
102 |
+
logger.info(f"Error processing {row_id}: {e}")
|
103 |
+
return None
|
104 |
+
|
105 |
+
combined = list(zip(sequences, modalities))
|
106 |
+
combined.sort(key=lambda x: x[1] != "polypeptide(L)")
|
107 |
+
sequences, modalities = zip(*combined) if combined else ([], [])
|
108 |
+
|
109 |
+
sequences = (list(sequences) + [None] * 5)[:5]
|
110 |
+
modalities = (list(modalities) + [None] * 5)[:5]
|
111 |
+
|
112 |
+
return {
|
113 |
+
"id": row_id,
|
114 |
+
**{f"sequence{i + 1}": sequences[i] for i in range(5)},
|
115 |
+
**{f"modality{i + 1}": modalities[i] for i in range(5)},
|
116 |
+
}
|
117 |
+
|
118 |
+
|
119 |
+
def process_rows(df: pd.DataFrame, sequence_extractor: callable, debug: bool = False) -> pd.DataFrame:
|
120 |
+
"""
|
121 |
+
Process rows in the dataframe to extract sequence and modality information.
|
122 |
+
|
123 |
+
Parameters
|
124 |
+
----------
|
125 |
+
df : pd.DataFrame
|
126 |
+
Input dataframe with 'id' column.
|
127 |
+
sequence_extractor : callable
|
128 |
+
Function to extract sequences and modalities given a row id.
|
129 |
+
debug : bool, optional
|
130 |
+
If True, only processes first 10 rows.
|
131 |
+
|
132 |
+
Returns
|
133 |
+
-------
|
134 |
+
pd.DataFrame
|
135 |
+
Dataframe with sequence and modality columns.
|
136 |
+
"""
|
137 |
+
row_ids = df["id"].tolist()[:10] if debug else df["id"].tolist()
|
138 |
+
results = []
|
139 |
+
with ThreadPoolExecutor(max_workers=MAX_WORKERS) as executor:
|
140 |
+
futures = {executor.submit(_safe_extract, rid, sequence_extractor): rid for rid in row_ids}
|
141 |
+
for future in tqdm(as_completed(futures), total=len(futures)):
|
142 |
+
result = future.result()
|
143 |
+
if result:
|
144 |
+
results.append(result)
|
145 |
+
return pd.DataFrame(results).replace(MODALITY_MAPPING)
|
146 |
+
|
147 |
+
|
148 |
+
|
149 |
+
def extract_protein_nucleotide(row_id: str) -> tuple[list[str], list[str]]:
|
150 |
+
"""Example ID: 3adi_1.pdb_3adi_1_RNA_D&E.pdb"""
|
151 |
+
pdb_id = row_id.split("_")[0]
|
152 |
+
pdb_data = _fetch_pdb_data(pdb_id)
|
153 |
+
sequences, types = (
|
154 |
+
zip(*[(m["sequence"], m["molecule_type"]) for m in pdb_data if "sequence" in m]) if pdb_data else ([], [])
|
155 |
+
)
|
156 |
+
return list(sequences), list(types)
|
157 |
+
|
158 |
+
|
159 |
+
def extract_protein_small_molecule(row_id: str) -> tuple[list[str], list[str]]:
|
160 |
+
"""Example ID: 4h4e_1.pdb_4h4e_1_10G_D.pdb"""
|
161 |
+
parts = row_id.split("_")
|
162 |
+
pdb_id, ligand_id = parts[2], parts[4]
|
163 |
+
pdb_data = _fetch_pdb_data(pdb_id)
|
164 |
+
sequences, types = (
|
165 |
+
zip(*[(m["sequence"], m["molecule_type"]) for m in pdb_data if "sequence" in m]) if pdb_data else ([], [])
|
166 |
+
)
|
167 |
+
ligand = _fetch_smiles_data(ligand_id)
|
168 |
+
if ligand:
|
169 |
+
sequences += (ligand,)
|
170 |
+
types += ("smiles",)
|
171 |
+
return list(sequences), list(types)
|
172 |
+
|
173 |
+
|
174 |
+
def extract_interacting_chains(row_id: str) -> tuple[list[str], list[str]]:
|
175 |
+
"""Example ID: 2uxq_2_A_B"""
|
176 |
+
pdb_id, _, chain_a, chain_b = row_id.split("_")
|
177 |
+
chains_of_interest = {chain_a, chain_b}
|
178 |
+
molecules = _fetch_pdb_data(pdb_id)
|
179 |
+
chain_map = {}
|
180 |
+
for m in molecules:
|
181 |
+
if "sequence" not in m or "in_chains" not in m:
|
182 |
+
continue
|
183 |
+
for chain in m["in_chains"]:
|
184 |
+
if chain in chains_of_interest:
|
185 |
+
chain_map[chain] = (m["sequence"], m["molecule_type"])
|
186 |
+
sequences, types = [], []
|
187 |
+
for chain in [chain_a, chain_b]:
|
188 |
+
if chain in chain_map:
|
189 |
+
seq, mol_type = chain_map[chain]
|
190 |
+
sequences.append(seq)
|
191 |
+
types.append(mol_type)
|
192 |
+
else:
|
193 |
+
sequences.append(None)
|
194 |
+
types.append(None)
|
195 |
+
return sequences, types
|
196 |
+
|
197 |
+
|
198 |
+
def process_protein_nucleotide(debug: bool = False) -> None:
|
199 |
+
df = pd.read_csv(f"{S3_RAW}/protein-dna.csv", sep="\t")
|
200 |
+
df_out = process_rows(df, extract_protein_nucleotide, debug=debug)
|
201 |
+
df_out = df_out.drop_duplicates().reset_index(drop=True)
|
202 |
+
df_out.to_parquet(f"{S3_PROCESSED}/protein-dna.parquet", index=False)
|
203 |
+
|
204 |
+
df = pd.read_csv(f"{S3_RAW}/protein-rna.csv", sep="\t")
|
205 |
+
df_out = process_rows(df, extract_protein_nucleotide, debug=debug)
|
206 |
+
df_out = df_out.drop_duplicates().reset_index(drop=True)
|
207 |
+
df_out.to_parquet(f"{S3_PROCESSED}/protein-rna.parquet", index=False)
|
208 |
+
|
209 |
+
|
210 |
+
def process_protein_protein(debug: bool = False) -> None:
|
211 |
+
df = pd.read_csv(f"{S3_RAW}/protein-protein.csv", sep="\t")
|
212 |
+
df_out = process_rows(df, extract_interacting_chains, debug=debug)
|
213 |
+
df_out = df_out.drop_duplicates().reset_index(drop=True)
|
214 |
+
df_out.to_parquet(f"{S3_PROCESSED}/protein-protein.parquet", index=False)
|
215 |
+
|
216 |
+
|
217 |
+
def process_protein_small_molecule(debug: bool = False) -> None:
|
218 |
+
df = pd.read_csv(f"{S3_RAW}/protein-small_molecule.csv", sep="\t")
|
219 |
+
df_out = process_rows(df, extract_protein_small_molecule, debug=debug)
|
220 |
+
df_out = df_out.dropna(how="all", axis=1).drop_duplicates().reset_index(drop=True)
|
221 |
+
df_out.to_parquet(f"{S3_PROCESSED}/protein-small_molecule.parquet", index=False)
|
222 |
+
|
223 |
+
|
224 |
+
def process_smiles_smiles() -> None:
|
225 |
+
df = pd.read_csv(f"{S3_RAW}/small_molecule-small_molecule.csv")
|
226 |
+
df = df["id"].str.split("_", expand=True)
|
227 |
+
df.columns = ["id", "sequence1", "num_1", "sequence2", "num_2"]
|
228 |
+
df = df[df["sequence1"] != df["sequence2"]][["sequence1", "sequence2"]]
|
229 |
+
df.insert(1, "modality1", "smiles")
|
230 |
+
df.insert(3, "modality2", "smiles")
|
231 |
+
df = df.drop_duplicates().reset_index(drop=True)
|
232 |
+
df.to_parquet(f"{S3_PROCESSED}/small_molecule-small_molecule.parquet", index=False)
|
233 |
+
|
234 |
+
|
235 |
+
if __name__ == "__main__":
|
236 |
+
upload_raw_datasets_to_s3()
|
237 |
+
process_protein_nucleotide(debug=False)
|
238 |
+
process_protein_protein(debug=False)
|
239 |
+
process_protein_small_molecule(debug=False)
|
240 |
+
process_smiles_smiles()
|