APPLICATIONS OF ARTIFICIAL INTELLIGENCE IN HUMAN GENOMIC RESEARCH AND PRECISION MEDICINE

Abstract

Background: The volume of human genomic data generated by whole-genome sequencing, genome-wide association studies (GWAS), and multi-omics platforms now far exceeds the interpretive capacity of conventional bioinformatic pipelines. Artificial intelligence (AI) and machine learning (ML) methods have emerged as essential tools for extracting clinically and biologically meaningful signals from these datasets. Main Findings: This review critically examines four major application domains: (1) pathogenic variant interpretation and functional annotation, where models such as AlphaFold2 and ACMG/AMP-guided ML classifiers have substantially improved resolution of variants of uncertain significance (VUS); (2) disease prediction and polygenic risk scoring, particularly for cancer and complex multifactorial conditions; (3) drug discovery and pharmacogenomics, including AI-driven target identification and personalised therapeutic selection; and (4) deep learning architectures in genomics, encompassing transformer-based sequence models, genomic foundation models such as the Nucleotide Transformer and DNABERT-2, and multi-omics integration frameworks. Significance: Despite compelling advances, critical barriers impede clinical translation: algorithmic bias arising from the historical underrepresentation of non-European ancestries in training datasets, limited model explainability, data privacy constraints, and fragmented regulatory frameworks. Federated learning and explainable AI (XAI) strategies are discussed as promising mitigation approaches. This review argues that realising the full potential of AI in precision genomic medicine requires not only technical innovation but also deliberate efforts towards equitable data collection and transparent model development.