code/scripts/sample_syngenome_negatives.py

"""Sample ~8,000 NON-AMR records from SynGenome with full ORFs for AMR-probe
validation negatives. Stratified across the four non-AMR functional classes.

Less-careful match than the VFDB pipeline (no length/GC pairing, no organism
matching) — these are validation negatives, not training negatives. Goal is
just "diverse non-AMR sequences with a real CDS inside", in the same JSONL
schema as the AMR positives so the embed pipeline picks them up.

Output: data/targeted_jsonl/syngenome_neg/<class_slug>.jsonl
        Each record's mag_id = "neg_<functional_class_slug>" so the embed
        path layout segregates negatives from positives.
"""
import argparse
import json
import random
import re
from collections import Counter, defaultdict
from pathlib import Path

# Reuse the ORF-finder from the AMR sampler
import sys
sys.path.insert(0, str(Path(__file__).parent))
from sample_syngenome_amr import has_full_orf, gc_content, class_slug, parse_score


def main():
    ap = argparse.ArgumentParser()
    ap.add_argument("--in-json", type=Path,
                    default=Path("/home/ror25cal/MGnify/syngenome_db.json"))
    ap.add_argument("--out-dir", type=Path,
                    default=Path("/home/ror25cal/MGnify/data/targeted_jsonl/syngenome_neg"))
    ap.add_argument("--target-n", type=int, default=8000)
    ap.add_argument("--per-class-cap", type=int, default=2500,
                    help="cap per functional_class for the 3 large classes; small classes taken in full")
    ap.add_argument("--seed", type=int, default=42)
    args = ap.parse_args()

    args.out_dir.mkdir(parents=True, exist_ok=True)
    rng = random.Random(args.seed)

    print(f"Loading {args.in_json} ...")
    with open(args.in_json) as f:
        d = json.load(f)
    non_amr = [r for r in d if r.get("functional_class") != "AMR"]
    print(f"  non-AMR records: {len(non_amr)}")

    print("Filtering to full-ORF records (start + stop, ≥50 aa) ...")
    full_orf = [r for r in non_amr if has_full_orf(r["sequence"].upper())]
    print(f"  {len(full_orf)} pass ({100*len(full_orf)/len(non_amr):.1f}%)")

    by_class: dict[str, list[dict]] = defaultdict(list)
    for r in full_orf:
        by_class[r.get("functional_class") or "UNKNOWN"].append(r)
    print("  available per functional_class:")
    for c, rs in sorted(by_class.items(), key=lambda kv: -len(kv[1])):
        print(f"    {c}: {len(rs)}")

    # Stratified pick
    chosen: list[dict] = []
    print("\nSampling:")
    for c, rs in sorted(by_class.items(), key=lambda kv: -len(kv[1])):
        n_take = min(len(rs), args.per_class_cap)
        sampled = rng.sample(rs, n_take) if n_take < len(rs) else rs
        chosen.extend(sampled)
        print(f"  {c}: taking {n_take} of {len(rs)}")
    print(f"\nTotal chosen: {len(chosen)} records")

    # Write JSONL per functional_class with negative-specific schema
    by_path: dict[str, list[dict]] = defaultdict(list)
    for r in chosen:
        seq = r["sequence"].upper()
        fc = r.get("functional_class") or "UNKNOWN"
        slug = f"neg_{class_slug(fc)}"
        rec = {
            "region_id": f"{r['seq_hash']}_NEG_SYN",
            "is_positive": False,
            "label": "negative",
            "label_class": fc,                     # e.g. "virulence", "stress", "mobile_element"
            "label_subclass": r.get("primary_label"),
            "secondary_label": r.get("secondary_label"),
            "gene_symbol": r.get("gene_name"),
            "product_name": r.get("product_name"),
            "organism": r.get("organism"),
            "source_db": "SynGenome",
            "source_accession": r.get("source_accession"),
            "evidence": r.get("evidence"),
            "evo2_score": parse_score(r.get("description", "")),
            "functional_class": fc,
            "primary_label": r.get("primary_label"),
            "pmids": r.get("pmids"),
            "cds_length": len(seq),
            "gc_content": round(gc_content(seq), 4),
            "mag_id": slug,                        # output-path grouping key
            "random_seed": args.seed,
            "sequence": seq,
        }
        clean = {k: (None if (isinstance(v, float) and v != v) else v)
                  for k, v in rec.items()}
        by_path[slug].append(clean)

    summary = []
    for slug, recs in sorted(by_path.items()):
        out_path = args.out_dir / f"{slug}.jsonl"
        with open(out_path, "w") as f:
            for r in recs:
                f.write(json.dumps(r) + "\n")
        summary.append((slug, len(recs)))
        print(f"  → {out_path.name}  ({len(recs)} records)")

    print(f"\nTotal: {sum(n for _, n in summary)} records across {len(summary)} files")
    summary_path = args.out_dir / "_sample_summary.csv"
    with open(summary_path, "w") as f:
        f.write("class_slug,n_records\n")
        for slug, n in summary:
            f.write(f"{slug},{n}\n")
    print(f"Summary: {summary_path}")


if __name__ == "__main__":
    main()