pipeline/prepare_three_new_sets.py

from __future__ import annotations

import json
import sys
import importlib.util
from pathlib import Path
from typing import Any

ROOT_DIR = Path(__file__).resolve().parents[1]
if str(ROOT_DIR) not in sys.path:
    sys.path.insert(0, str(ROOT_DIR))

from libs.utils.logging_utils import get_logger


def _load_large_library_builder():
    mod_path = ROOT_DIR / "libs/benchmark/large_library.py"
    spec = importlib.util.spec_from_file_location("large_library_runtime", mod_path)
    if spec is None or spec.loader is None:
        raise RuntimeError(f"Cannot load module spec from {mod_path}")
    mod = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(mod)
    return mod.build_large_benchmark_library


def _cfg(
    *,
    output_dir: str,
    docking_target_path: str,
    reference_id: str,
    pdb_id: str,
    ligand_comp_id: str,
    reference_smiles: str,
    chembl_target_id: str,
    shuffle_seed: int,
) -> dict[str, Any]:
    return {
        "benchmark_dataset": {
            "output_dir": output_dir,
            "target_size": 1000,
            "min_similarity_keep": 0.30,
            "pubchem_max_records": 12000,
            "pubchem_thresholds": [95, 90, 85, 80, 75, 70, 65],
            "chembl_target_id": chembl_target_id,
            "chembl_max_rows": 12000,
            "allow_generated_fallback": True,
            "reuse_existing": False,
            "shuffle_seed": int(shuffle_seed),
            "reference_smiles": reference_smiles,
        },
        "target": {
            "docking_target_path": docking_target_path,
        },
        "reference": {
            "reference_id": reference_id,
            "pdb_id": pdb_id,
            "ligand_comp_id": ligand_comp_id,
            "reference_smiles": reference_smiles,
        },
    }


def prepare_three_new_sets() -> dict[str, Any]:
    logger = get_logger("prepare_three_new_sets")
    build_large_benchmark_library = _load_large_library_builder()
    specs = [
        _cfg(
            output_dir="data/ligands/prelim_set_egfr_4wkq",
            docking_target_path="data/targets/prelim_set_egfr_4wkq/egfr_4wkq.pdb",
            reference_id="ref_gefitinib_prelim",
            pdb_id="4WKQ",
            ligand_comp_id="IRE",
            reference_smiles="COc1cc2ncnc(Nc3ccc(F)c(Cl)c3)c2cc1OCCCN1CCOCC1",
            chembl_target_id="CHEMBL203",
            shuffle_seed=20260423,
        ),
        _cfg(
            output_dir="data/ligands/prelim_set_abl1_1iep",
            docking_target_path="data/targets/prelim_set_abl1_1iep/abl1_1iep.pdb",
            reference_id="ref_imatinib_prelim",
            pdb_id="1IEP",
            ligand_comp_id="STI",
            reference_smiles="Cc1ccc(NC(=O)c2ccc(CN3CCN(C)CC3)cc2)cc1Nc1nccc(-c2cccnc2)n1",
            chembl_target_id="CHEMBL1862",
            shuffle_seed=20260424,
        ),
        _cfg(
            output_dir="data/ligands/prelim_set_mdm2_4hg7",
            docking_target_path="data/targets/prelim_set_mdm2_4hg7/mdm2_4hg7.pdb",
            reference_id="ref_nutlin3a_prelim",
            pdb_id="4HG7",
            ligand_comp_id="NUT",
            reference_smiles="Cc1nc2ccccc2n1CC(C)(C)c1cc(C(F)(F)F)cc(C(F)(F)F)c1",
            chembl_target_id="CHEMBL5023",
            shuffle_seed=20260425,
        ),
    ]

    payload: list[dict[str, Any]] = []
    for cfg in specs:
        out = build_large_benchmark_library(cfg, ROOT_DIR, logger)
        ref = out["reference_df"].iloc[0].to_dict()
        payload.append(
            {
                "ligands_dir": cfg["benchmark_dataset"]["output_dir"],
                "target_path": cfg["target"]["docking_target_path"],
                "shared_library": str(Path(cfg["benchmark_dataset"]["output_dir"]) / "shared_library_shuffled.csv"),
                "reference_csv": str(Path(cfg["benchmark_dataset"]["output_dir"]) / "reference_ligands.csv"),
                "reference_id": str(ref["reference_id"]),
                "reference_comp_id": str(ref["ligand_comp_id"]),
                "library_size": int(out["shuffled_df"].shape[0]),
            }
        )
        logger.info("Prepared %s size=%s", cfg["benchmark_dataset"]["output_dir"], out["shuffled_df"].shape[0])

    manifest = {
        "run_name": "prepare_three_new_sets",
        "datasets": payload,
    }
    out_manifest = ROOT_DIR / "data/ligands/prelim_three_sets_manifest.json"
    out_manifest.write_text(json.dumps(manifest, indent=2), encoding="utf-8")
    return manifest


if __name__ == "__main__":
    print(json.dumps(prepare_three_new_sets(), indent=2))