scripts/04_validate_resource.py

"""Validate core files for the PD discovery benchmark repository.

The checks are intentionally lightweight so they can run in GitHub Actions
without external biomedical API access.
"""

from __future__ import annotations

from pathlib import Path

import pandas as pd


ROOT = Path(__file__).resolve().parents[1]

REQUIRED_FILES = [
    "README.md",
    "DATASET_CARD.md",
    "CITATION.cff",
    "LICENSE",
    "dataset-metadata.json",
    "dashboard/app.py",
    "data/pd_discovery_target_benchmark.csv",
    "data/compound_selectivity_safety_matrix.csv",
    "data/validated_repurposing_candidates.csv",
    "data/do_not_prioritise_or_comparator_compounds.csv",
    "data/experimental_validation_matrix.csv",
    "data/pd_discovery_benchmark_knowledge_graph.graphml",
    "data/omics_recurrence/pd_multi_dataset_pathway_recurrence.csv",
    "figures/benchmark_target_ranking.png",
    "figures/benchmark_evidence_matrix.png",
    "figures/compound_selectivity_safety_triage.png",
    "reports/resource_manuscript_target_to_intervention_benchmark.md",
    "reproducibility/finalization_claim_audit.csv",
]


def require(condition: bool, message: str) -> None:
    if not condition:
        raise SystemExit(message)


def main() -> None:
    missing = [path for path in REQUIRED_FILES if not (ROOT / path).exists()]
    require(not missing, f"Missing required files: {missing}")

    targets = pd.read_csv(ROOT / "data/pd_discovery_target_benchmark.csv")
    require({"symbol", "benchmark_consensus_score_0_100", "benchmark_label"}.issubset(targets.columns), "Target benchmark columns are incomplete")
    require(len(targets) >= 5, "Target benchmark has too few rows")
    require(targets["benchmark_consensus_score_0_100"].between(0, 100).all(), "Target scores must be 0-100")

    validated = pd.read_csv(ROOT / "data/validated_repurposing_candidates.csv")
    require(len(validated) >= 3, "Validated candidate table has too few rows")
    require("curation_class" in validated.columns, "Validated candidate table lacks curation_class")

    exclusions = pd.read_csv(ROOT / "data/do_not_prioritise_or_comparator_compounds.csv")
    require(len(exclusions) >= 10, "Comparator/exclusion table has too few rows")

    audit = pd.read_csv(ROOT / "reproducibility/finalization_claim_audit.csv")
    require("clinical guardrail" in set(audit["domain"]), "Claim audit must include clinical guardrail")
    require(audit["status"].astype(str).str.lower().eq("pass").all(), "One or more claim-audit checks failed")

    print("PD discovery benchmark validation passed.")


if __name__ == "__main__":
    main()