src/validation/interaction_sensitivity.py

import numpy as np, pandas as pd
from pathlib import Path
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LinearRegression

DATA = Path("data/processed"); RES = Path("results"); RES.mkdir(exist_ok=True, parents=True)
C = pd.read_csv(DATA/"causal_table.csv")

anchors = ["PDR5_expr","SNQ2_expr","YOR1_expr","ATM1_expr"]
regs = ["YAP1_reg","PDR1_reg"]

rows=[]
for t in anchors:
    df = C[[t,"outcome"]+regs].dropna().copy()
    for r in regs:
        Z = df[[t, r]].astype(float).values
        Z = Z - Z.mean(0, keepdims=True)
        T = Z[:,0:1]; R = Z[:,1:2]
        X = np.hstack([T, R, T*R])
        y = df["outcome"].astype(float).values
        nuis = ["ethanol_pct","ROS","NaCl_mM","H2O2_uM"]
        W = df.join(C[nuis], how="left")[nuis].astype(float).values
        W = W - W.mean(0, keepdims=True)
        Xfull = np.hstack([X, W])
        mdl = LinearRegression().fit(Xfull, y)
        beta_T, beta_R, beta_TR = mdl.coef_[0], mdl.coef_[1], mdl.coef_[2]
        rows.append([t.replace("_expr",""), r, beta_T, beta_TR])

INT = pd.DataFrame(rows, columns=["transporter","regulator","main_T","interaction_TxReg"])
INT.to_csv(RES/"validation_interactions.csv", index=False)
print(" Saved:", RES/"validation_interactions.csv")
display(INT.sort_values(["transporter","regulator"]))