Active Causal Hypothesis Testing for AI-Guided Drug Target Discovery

David Scott Lewis, Enrique Zueco

Keywords: Active Causal Hypothesis Testing (ACHT), AI-guided drug discovery, virtual experimenter, differentiable causal discovery (NOTEARS), graph neural networks (GNNs), Graph Attention Network (GAT), Bayesian active learning, Expected Conditional Improvement (ECI), information gain prioritization, Monte Carlo Wavelet Coherence, causal DAG, mechanistic hypotheses

TL;DR: ACHT is an AI “virtual experimenter” that fuses GNNs, differentiable causal discovery, and Bayesian active learning to rank causal drug targets in STRING, surfacing 3,529 candidates and validating mechanisms via Monte Carlo wavelet coherence.

Submission Number: 51