src/validation/lit_crosscheck.py

import pandas as pd

causal_hits = pd.read_csv("results/fig_SIMS_waterfall.csv") if (Path("results/fig_SIMS_waterfall.csv").exists()) else None

landmark_refs = {
    "PDR5": "The major yeast multidrug transporter; confers resistance to azoles, cycloheximide, ethanol tolerance (Paulsen et al., Trends Microbiol 1996; Piper et al., Appl Environ Microbiol 1998).",
    "SNQ2": "Broad-spectrum drug resistance; deletion leads to hypersensitivity to oxidants, ethanol, and xenobiotics (Decottignies et al., EMBO J 1995).",
    "YOR1": "Oligomycin and multiple toxin resistance, associated with ATPase-dependent efflux (Katzmann et al., Mol Cell Biol 1995).",
    "ATM1": "Mitochondrial ABC transporter, iron-sulfur cluster export, not classically ethanol but key for oxidative stress resilience (Kispal et al., EMBO J 1999)."
}

rows = []
for t, ref in landmark_refs.items():
    rows.append({"transporter": t, "supporting_evidence": ref})
lit_check = pd.DataFrame(rows)
lit_check.to_csv("results/validation_literature_crosscheck.csv", index=False)
lit_check

import json, pandas as pd, numpy as np
from pathlib import Path

RES = Path("results"); RES.mkdir(exist_ok=True, parents=True)

s3_path = RES/"causal_section3_snapshot.json"
with open(s3_path, "r") as f:
    s3 = json.load(f)

def _to_scalar(x):
    """Return a float scalar from x, handling dicts like {'overall': v, ...} or {'ATE': v}."""
    if x is None:
        return np.nan
    if isinstance(x, (int, float, np.number)):
        return float(x)
    if isinstance(x, dict):
        for k in ["overall", "ATE", "ate", "mean", "Mean", "val", "value"]:
            if k in x and isinstance(x[k], (int, float, np.number)):
                return float(x[k])
        for v in x.values():
            if isinstance(v, (int, float, np.number)):
                return float(v)
        return np.nan
    if isinstance(x, (list, tuple)):
        for v in x:
            if isinstance(v, (int, float, np.number)):
                return float(v)
        return np.nan
    return np.nan

ATE_table = s3.get("ATE_table") or {}
if not ATE_table and "stress_ate" in s3:
    ATE_table = {k: v.get("overall", np.nan) for k, v in s3["stress_ate"].items() if isinstance(v, dict)}

anchors = {
    "PDR5_expr": ("PDR5", "Ethanol/drug tolerance via efflux; KO sensitizes", "positive"),
    "SNQ2_expr": ("SNQ2", "Oxidant & ethanol hypersensitivity when deleted", "positive"),
    "YOR1_expr": ("YOR1", "Drug/toxin efflux", "positive or context-specific"),
    "ATM1_expr": ("ATM1", "Mitochondrial Fe–S export; oxidative resilience", "positive under oxidative"),
}

rows=[]
for key,(gene,phen,expect) in anchors.items():
    val = _to_scalar(ATE_table.get(key, np.nan))
    if pd.notna(val):
        sgn = "positive" if val > 0 else ("negative" if val < 0 else "~0")
    else:
        sgn = "~0"
    note = (
        "matches expected direction"
        if (expect.startswith("positive") and pd.notna(val) and val > 0)
        or ("context" in expect and pd.notna(val) and val != 0)
        else ("opposite sign (check stress/context)" if pd.notna(val) and val != 0 else "no effect / not detected")
    )
    rows.append([gene, phen, expect, (float(val) if pd.notna(val) else np.nan), 3, sgn, note])

df = pd.DataFrame(rows, columns=["gene","phenotype","expected_sign","mean_ATE","n_stresses","ATE_sign","concordance_note"])
out = RES/"validation_lit_crosscheck.csv"
df.to_csv(out, index=False)
print(" Saved:", out)
display(df)

import json, numpy as np, pandas as pd, matplotlib.pyplot as plt, seaborn as sns
from pathlib import Path

RES = Path("results"); RES.mkdir(parents=True, exist_ok=True)

# helpers
def _to_scalar(x):
    if x is None: return np.nan
    if isinstance(x, (int,float,np.number)): return float(x)
    if isinstance(x, dict):
        for k in ["overall","ATE","ate","mean","Mean","val","value"]:
            if k in x and isinstance(x[k], (int,float,np.number)): return float(x[k])
        # first numeric value
        for v in x.values():
            if isinstance(v, (int,float,np.number)): return float(v)
        return np.nan
    if isinstance(x, (list,tuple)):
        for v in x:
            if isinstance(v, (int,float,np.number)): return float(v)
        return np.nan
    return np.nan

with open(RES/"causal_section3_snapshot.json","r") as f:
    s3 = json.load(f)

anchors = {
    "PDR5_expr": "PDR5",
    "SNQ2_expr": "SNQ2",
    "YOR1_expr": "YOR1",
    "ATM1_expr": "ATM1",
}

rows = []

if "stress_ate" in s3:
    stab = s3["stress_ate"] 
    def _looks_like_transporter_keys(d):
        ks = list(d.keys())[:5]
        return any(isinstance(k,str) and k.endswith("_expr") for k in ks)
    if isinstance(stab, dict) and stab:
        first = next(iter(stab.values()))
        if isinstance(first, dict) and _looks_like_transporter_keys(stab):  
            for t, inner in stab.items():
                if t not in anchors: continue
                for stress, val in inner.items():
                    v = _to_scalar(val)
                    if pd.notna(v): rows.append({"transporter": t, "stress": str(stress), "ATE": v})
        elif isinstance(first, dict):  
            for stress, inner in stab.items():
                for t, val in inner.items():
                    if t not in anchors: continue
                    v = _to_scalar(val)
                    if pd.notna(v): rows.append({"transporter": t, "stress": str(stress), "ATE": v})

if not rows:
    A = s3.get("ATE_table", {})
    for t, g in anchors.items():
        v = _to_scalar(A.get(t, np.nan))
        if pd.notna(v):
            rows.append({"transporter": t, "stress": "overall", "ATE": v})

S = pd.DataFrame(rows)

out_csv = RES/"validation_anchor_per_stress.csv"
S.to_csv(out_csv, index=False)
print(" Saved:", out_csv)
display(S.head(10))

if not S.empty and S["ATE"].notna().any():
    pt = S.pivot_table(index="transporter", columns="stress", values="ATE", aggfunc="mean")
    pt = pt.dropna(axis=0, how="all").dropna(axis=1, how="all")
    if pt.size > 0:
        plt.figure(figsize=(max(6, 1.8 + 0.6*pt.shape[1]), max(2.8, 0.45*pt.shape[0])))
        sns.heatmap(pt, annot=True, fmt=".3f", center=0, cmap="coolwarm")
        plt.title("Per-stress ATEs for anchor genes")
        plt.tight_layout()
        plt.savefig(RES/"validation_anchor_per_stress_heatmap.png", dpi=300)
        plt.show()
        print(" Saved heatmap →", RES/"validation_anchor_per_stress_heatmap.png")
    else:
        print(" No non-NaN cells after pivot; skipping heatmap.")
else:
    print("No per-stress ATEs available; heatmap skipped.")