Explainable AI in Genomic Medicine

Abstract

Artificial intelligence is transforming genomic medicine, achieving state-of-the-art performance in variant pathogenicity prediction, polygenic risk scoring, clinical trial matching, and therapeutic recommendation. Yet clinical adoption remains constrained by a fundamental concern: deep learning models behave as black boxes that cannot explain why they make specific predictions, limiting clinician trust, regulatory approval, and patient autonomy in healthcare decisions. Explainable AI (XAI) techniques -- attention visualisation, SHAP values, counterfactual explanations, mechanistic interpretability, and causal inference -- aim to render model reasoning transparent while preserving predictive performance. The challenge in genomic medicine is particularly acute: explanations must be biologically meaningful (connecting to genes, pathways, molecular mechanisms), clinically actionable (supporting therapeutic decisions), and legally defensible (satisfying FDA and EU AI Act transparency requirements). We present the Explainable Genomic AI Framework (EGAIF), evaluating five XAI approaches across four genomic medicine applications (variant pathogenicity classification, polygenic risk score interpretation, drug response prediction, and clinical trial matching). Our Explainable Genomic AI Score (EGAS) measures explanation fidelity to model behaviour, biological plausibility, clinical utility, user comprehension, and regulatory compliance. Mechanistic interpretability achieves the highest EGAS (0.926) through direct mapping of neural network internals to biological pathways, while counterfactual explanations achieve the highest clinical actionability (0.960) by showing clinicians which genetic changes would alter predictions