Create scorer.py

scorer.py

@@ -0,0 +1,222 @@
+from __future__ import annotations
+
+import json
+from dataclasses import dataclass
+from typing import Dict, List
+
+import pandas as pd
+
+
+REQUIRED_COLS = [
+    "row_id",
+    "series_id",
+    "inoculum_rank",
+    "organism",
+    "strain_id",
+    "antibiotic_name",
+    "antibiotic_class",
+    "exposure_index",
+    "mic_mg_L",
+    "cfu0_log10",
+    "cfu24_log10",
+    "kill_24_log10",
+    "media",
+    "assay_method",
+    "source_type",
+    "inoculum_effect_signal",
+    "earliest_inoculum_effect",
+]
+
+
+@dataclass
+class Thresholds:
+    min_points: int = 3
+
+    # event trigger
+    inoculum_rise_min: float = 1.0          # log10 units vs baseline
+    kill_drop_min: float = 0.8              # drop in kill vs baseline
+
+    # gates to avoid confounds
+    exposure_fold_min: float = 0.90
+    mic_fold_max: float = 2.0
+
+
+def _validate(df: pd.DataFrame) -> List[str]:
+    errs: List[str] = []
+
+    missing = [c for c in REQUIRED_COLS if c not in df.columns]
+    if missing:
+        errs.append(f"missing_columns: {missing}")
+
+    for c in ["inoculum_rank", "exposure_index", "mic_mg_L", "cfu0_log10", "cfu24_log10", "kill_24_log10"]:
+        if c in df.columns and df[c].isna().any():
+            errs.append(f"null_values_in: {c}")
+
+    for c in ["exposure_index", "mic_mg_L"]:
+        if c in df.columns:
+            bad = (df[c] <= 0).sum()
+            if bad:
+                errs.append(f"non_positive_values_in: {c} count={int(bad)}")
+
+    for c in ["inoculum_effect_signal", "earliest_inoculum_effect"]:
+        if c in df.columns:
+            bad = (~df[c].isin([0, 1])).sum()
+            if bad:
+                errs.append(f"non_binary_values_in: {c} count={int(bad)}")
+
+    counts = df.groupby("series_id")["earliest_inoculum_effect"].sum()
+    bad_series = counts[counts > 1].index.tolist()
+    if bad_series:
+        errs.append(f"multiple_earliest_inoculum_effect_in_series: {bad_series}")
+
+    return errs
+
+
+def _f1(tp: int, fp: int, fn: int) -> float:
+    denom = 2 * tp + fp + fn
+    return 0.0 if denom == 0 else (2 * tp) / denom
+
+
+def score(path: str) -> Dict[str, object]:
+    df = pd.read_csv(path)
+    errors = _validate(df)
+    if errors:
+        return {"ok": False, "errors": errors}
+
+    t = Thresholds()
+
+    df = df.sort_values(["series_id", "inoculum_rank"]).reset_index(drop=True)
+    df["pred_earliest_inoculum_effect"] = 0
+    df["pred_inoculum_effect_signal"] = 0
+    df["flag_exposure_drop"] = 0
+    df["flag_mic_shift"] = 0
+
+    series_rows: List[Dict[str, object]] = []
+
+    for sid, g in df.groupby("series_id"):
+        g = g.sort_values("inoculum_rank").copy()
+
+        if len(g) < t.min_points:
+            series_rows.append(
+                {
+                    "series_id": sid,
+                    "y_effect": int(g["inoculum_effect_signal"].max()),
+                    "p_effect": 0,
+                    "true_transition_row_id": (str(g[g["earliest_inoculum_effect"] == 1].iloc[0]["row_id"]) if (g["earliest_inoculum_effect"] == 1).any() else None),
+                    "pred_transition_row_id": None,
+                    "flags": ["too_few_points"],
+                }
+            )
+            continue
+
+        base = g.iloc[0]
+        base_inoc = float(base["cfu0_log10"])
+        base_kill = float(base["kill_24_log10"])
+        base_exp = float(base["exposure_index"])
+        base_mic = float(base["mic_mg_L"])
+
+        hits = []
+        for idx, row in g.iterrows():
+            inoc = float(row["cfu0_log10"])
+            kill = float(row["kill_24_log10"])
+            exp = float(row["exposure_index"])
+            mic = float(row["mic_mg_L"])
+
+            exp_fold = exp / base_exp if base_exp > 0 else 0.0
+            mic_fold = mic / base_mic if base_mic > 0 else 99.0
+
+            if exp_fold < t.exposure_fold_min:
+                df.loc[idx, "flag_exposure_drop"] = 1
+            if mic_fold > t.mic_fold_max:
+                df.loc[idx, "flag_mic_shift"] = 1
+
+            if idx == g.index[0]:
+                continue
+
+            inoc_rise = inoc - base_inoc
+            kill_drop = base_kill - kill
+
+            candidate = (
+                inoc_rise >= t.inoculum_rise_min
+                and kill_drop >= t.kill_drop_min
+                and exp_fold >= t.exposure_fold_min
+                and mic_fold <= t.mic_fold_max
+            )
+
+            if candidate:
+                hits.append(idx)
+
+        if hits:
+            first = hits[0]
+            df.loc[first, "pred_earliest_inoculum_effect"] = 1
+            later = g[g.index >= first].index
+            df.loc[later, "pred_inoculum_effect_signal"] = 1
+
+        y = int(g["inoculum_effect_signal"].max())
+        p = int(df.loc[g.index, "pred_inoculum_effect_signal"].max())
+
+        true_tr = g[g["earliest_inoculum_effect"] == 1]
+        true_id = None
+        if len(true_tr) == 1:
+            true_id = str(true_tr.iloc[0]["row_id"])
+
+        pred_tr = None
+        pred_tr_rows = df.loc[g.index][df.loc[g.index, "pred_earliest_inoculum_effect"] == 1]
+        if len(pred_tr_rows) == 1:
+            pred_tr = str(pred_tr_rows.iloc[0]["row_id"])
+
+        series_rows.append(
+            {
+                "series_id": sid,
+                "y_effect": y,
+                "p_effect": p,
+                "true_transition_row_id": true_id,
+                "pred_transition_row_id": pred_tr,
+                "exposure_drop_flags": int(df.loc[g.index, "flag_exposure_drop"].sum()),
+                "mic_shift_flags": int(df.loc[g.index, "flag_mic_shift"].sum()),
+            }
+        )
+
+    sr = pd.DataFrame(series_rows)
+
+    tp = int(((sr["y_effect"] == 1) & (sr["p_effect"] == 1)).sum())
+    fp = int(((sr["y_effect"] == 0) & (sr["p_effect"] == 1)).sum())
+    fn = int(((sr["y_effect"] == 1) & (sr["p_effect"] == 0)).sum())
+    tn = int(((sr["y_effect"] == 0) & (sr["p_effect"] == 0)).sum())
+
+    transition_hit = int(
+        (
+            sr["true_transition_row_id"].notna()
+            & (sr["true_transition_row_id"] == sr["pred_transition_row_id"])
+        ).sum()
+    )
+    transition_miss = int(
+        (
+            sr["true_transition_row_id"].notna()
+            & (sr["true_transition_row_id"] != sr["pred_transition_row_id"])
+        ).sum()
+    )
+
+    return {
+        "ok": True,
+        "path": path,
+        "counts": {"tp": tp, "fp": fp, "fn": fn, "tn": tn},
+        "metrics": {
+            "f1_series": _f1(tp, fp, fn),
+            "transition_hit": transition_hit,
+            "transition_miss": transition_miss,
+            "n_series": int(len(sr)),
+        },
+        "series_table": series_rows,
+    }
+
+
+if __name__ == "__main__":
+    import argparse
+
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--path", required=True)
+    args = ap.parse_args()
+
+    result = score(args.path)
+    print(json.dumps(result, indent=2))