Artificial intelligence and multi-omics integration in Alzheimer’s disease: translational challenges

The past two decades have witnessed a revolution in high-throughput omics technologies, which now offer unprecedented views of the molecular complexity underlying Alzheimer’s Disease (AD) and associated biomarkers. By leveraging nonlinear models and pattern recognition, artificial intelligence (AI) can integrate omics data to transcend the limitations of individual modalities, improving early detection and molecular subtype identification, expediting the identification of possible therapies via the inference of molecular pathways, and tracking disease progression. Yet, despite these breakthroughs, most AI-omics applications remain confined to research settings. From this perspective, we focus on three critical translational challenges: algorithmic bias, interpretability and trust, and the implementation of AI-driven omics into clinical practice. This perspective advocates for inclusive data infrastructures, improving interpretability and trust by providing greater mechanistic insight, and improving clinical translation by increasing reproducibility and cost-effectiveness. Such lessons from AD will seek to inform translational medicine across diverse diseases, beyond AD.