Integration of multi-omics and artificial intelligence for therapeutic insights in Alzheimer's disease: A comprehensive review

Abstract

Alzheimer's disease (AD) is a progressively worsening type of brain disorder that damages the nerve cells. It is marked by the buildup of amyloid-β plaques outside the cells, tau neurofibrillary tangles inside the cells, and overall molecular-level dysfunction. The therapies currently available mainly cater to alleviating the symptoms, whereas the newly approved disease-modifying antibodies, such as lecanemab and donanemab, bring out only limited clinical improvements. Being complicated and involving many factors, AD requires sophisticated computer-based methods to combine different biological data and find suitable therapy targets. In this review, we discuss how artificial intelligence (AI)-powered multi-omics data integration can be a catalyst in discovering drug targets, identifying biomarkers, and stratifying patients for AD. By utilizing machine learning techniques like random forests, graph neural networks, and deep learning, AI-led multi-omics methods have helped uncover new therapeutic targets. Models that were built using federated learning across various institutions outperformed single-center models with a higher area under the curve score (0.84, 0.94 versus 0.76, 0.85). AI-guided patient stratification lessened the clinical trial's sample size needs by 40, 55% while still retaining 80, 90% statistical power. Multi-omics analyses further pointed out that it is the downstream molecular pathways, and not amyloid pathology alone, that are significantly involved in disease progression, thereby questioning the effectiveness of single-target anti-amyloid therapies and endorsing combination treatment strategies. AI and multi-omics data combination can be a game-changer in facilitating new target discovery, making clinical trial design more efficient, and ushering in precision medicine in AD.