Datasets:
Formats:
parquet
Languages:
English
Size:
10M - 100M
Tags:
biology
chemistry
drug-discovery
clinical-trials
protein-protein-interaction
gene-essentiality
License:
File size: 7,316 Bytes
6d1bbc7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 | #!/usr/bin/env python3
"""Prepare PPI ML benchmark datasets from exported negatives + HuRI positives.
Produces (in exports/ppi/):
- ppi_m1_balanced.parquet — 1:1 pos/neg (NegBioDB curated negatives)
- ppi_m1_realistic.parquet — 1:10 pos/neg (NegBioDB curated negatives)
- ppi_m1_uniform_random.parquet — 1:1 pos/neg (Exp 1 control A)
- ppi_m1_degree_matched.parquet — 1:1 pos/neg (Exp 1 control B)
- ppi_m1_balanced_ddb.parquet — M1 balanced + split_degree_balanced (Exp 4)
Prerequisite:
- exports/ppi/negbiodb_ppi_pairs.parquet (from export step)
- data/ppi/huri/HI-union.tsv
- data/ppi/huri/ensg_to_uniprot.tsv
- data/negbiodb_ppi.db
Usage:
PYTHONPATH=src python scripts_ppi/prepare_exp_data.py
"""
from __future__ import annotations
import argparse
import logging
import sys
from pathlib import Path
import pandas as pd
logging.basicConfig(
level=logging.INFO,
format="%(asctime)s %(levelname)s %(message)s",
datefmt="%H:%M:%S",
)
logger = logging.getLogger(__name__)
ROOT = Path(__file__).parent.parent
def main(argv: list[str] | None = None) -> int:
parser = argparse.ArgumentParser(description="Prepare PPI ML benchmark datasets")
parser.add_argument(
"--data-dir", type=Path, default=ROOT / "exports" / "ppi",
help="Directory for input/output parquets",
)
parser.add_argument(
"--db", type=Path, default=ROOT / "data" / "negbiodb_ppi.db",
help="Path to negbiodb_ppi.db",
)
parser.add_argument(
"--huri-dir", type=Path, default=ROOT / "data" / "ppi" / "huri",
help="Directory containing HI-union.tsv and ensg_to_uniprot.tsv",
)
parser.add_argument("--seed", type=int, default=42)
parser.add_argument("--skip-exp1", action="store_true")
parser.add_argument("--skip-exp4", action="store_true")
args = parser.parse_args(argv)
data_dir: Path = args.data_dir
db_path: Path = args.db
huri_dir: Path = args.huri_dir
# Verify inputs
neg_parquet = data_dir / "negbiodb_ppi_pairs.parquet"
huri_file = huri_dir / "HI-union.tsv"
ensg_file = huri_dir / "ensg_to_uniprot.tsv"
for p in [neg_parquet, huri_file, ensg_file, db_path]:
if not p.exists():
logger.error("Required file missing: %s", p)
return 1
data_dir.mkdir(parents=True, exist_ok=True)
# ------------------------------------------------------------------
# Load negatives + positives
# ------------------------------------------------------------------
from negbiodb_ppi.export import (
load_huri_positives_df,
resolve_conflicts,
build_m1_balanced,
build_m1_realistic,
add_degree_balanced_split,
apply_ppi_m1_splits,
generate_uniform_random_negatives,
generate_degree_matched_negatives,
control_pairs_to_df,
)
logger.info("Loading negatives from %s", neg_parquet.name)
neg_df = pd.read_parquet(neg_parquet)
logger.info("Negatives: %d rows", len(neg_df))
# Filter to rows with sequences
has_seq = neg_df["sequence_1"].notna() & neg_df["sequence_2"].notna()
neg_df = neg_df[has_seq].reset_index(drop=True)
logger.info("Negatives with sequences: %d rows", len(neg_df))
logger.info("Loading HuRI positives...")
pos_df = load_huri_positives_df(
ppi_path=huri_file,
ensg_mapping_path=ensg_file,
db_path=db_path,
)
logger.info("Positives: %d rows", len(pos_df))
# Conflict resolution
neg_df, pos_df, n_conflicts = resolve_conflicts(neg_df, pos_df)
logger.info("After conflict resolution: %d neg, %d pos, %d conflicts removed",
len(neg_df), len(pos_df), n_conflicts)
# Positive pair set for exclusion
positive_pairs = set(zip(pos_df["uniprot_id_1"], pos_df["uniprot_id_2"]))
# ------------------------------------------------------------------
# M1 datasets (NegBioDB curated negatives)
# ------------------------------------------------------------------
logger.info("Building M1 balanced (1:1)...")
m1_bal = build_m1_balanced(neg_df, pos_df, seed=args.seed)
m1_bal_path = data_dir / "ppi_m1_balanced.parquet"
m1_bal.to_parquet(m1_bal_path, index=False)
logger.info("Saved %d rows → %s", len(m1_bal), m1_bal_path.name)
logger.info("Building M1 realistic (1:10)...")
m1_real = build_m1_realistic(neg_df, pos_df, ratio=10, seed=args.seed)
m1_real_path = data_dir / "ppi_m1_realistic.parquet"
m1_real.to_parquet(m1_real_path, index=False)
logger.info("Saved %d rows → %s", len(m1_real), m1_real_path.name)
# ------------------------------------------------------------------
# Exp 1: Control negatives
# ------------------------------------------------------------------
if not args.skip_exp1:
n_pos = len(pos_df)
logger.info("Generating %d uniform random negatives (Exp 1A)...", n_pos)
uniform_pairs = generate_uniform_random_negatives(
db_path, positive_pairs, n_samples=n_pos, seed=args.seed
)
uniform_neg = control_pairs_to_df(uniform_pairs, db_path)
m1_uniform = pd.concat([pos_df, uniform_neg], ignore_index=True)
m1_uniform = apply_ppi_m1_splits(m1_uniform, seed=args.seed)
uniform_path = data_dir / "ppi_m1_uniform_random.parquet"
m1_uniform.to_parquet(uniform_path, index=False)
logger.info("Saved %d rows → %s", len(m1_uniform), uniform_path.name)
logger.info("Generating %d degree-matched negatives (Exp 1B)...", n_pos)
deg_pairs = generate_degree_matched_negatives(
db_path, positive_pairs, n_samples=n_pos, seed=args.seed
)
deg_neg = control_pairs_to_df(deg_pairs, db_path)
m1_deg = pd.concat([pos_df, deg_neg], ignore_index=True)
m1_deg = apply_ppi_m1_splits(m1_deg, seed=args.seed)
deg_path = data_dir / "ppi_m1_degree_matched.parquet"
m1_deg.to_parquet(deg_path, index=False)
logger.info("Saved %d rows → %s", len(m1_deg), deg_path.name)
else:
logger.info("Skipping Exp 1 (--skip-exp1)")
# ------------------------------------------------------------------
# Exp 4: DDB split
# ------------------------------------------------------------------
if not args.skip_exp4:
logger.info("Building M1 balanced + DDB split (Exp 4)...")
m1_ddb = add_degree_balanced_split(
pd.read_parquet(m1_bal_path), seed=args.seed
)
ddb_path = data_dir / "ppi_m1_balanced_ddb.parquet"
m1_ddb.to_parquet(ddb_path, index=False)
logger.info("Saved %d rows → %s", len(m1_ddb), ddb_path.name)
else:
logger.info("Skipping Exp 4 (--skip-exp4)")
# ------------------------------------------------------------------
# Summary
# ------------------------------------------------------------------
logger.info("=== Summary ===")
for f in sorted(data_dir.glob("ppi_m1_*.parquet")):
df = pd.read_parquet(f, columns=["Y"])
n_pos = (df["Y"] == 1).sum()
n_neg = (df["Y"] == 0).sum()
logger.info(" %s: %d rows (pos=%d, neg=%d)", f.name, len(df), n_pos, n_neg)
logger.info("Done.")
return 0
if __name__ == "__main__":
sys.exit(main())
|