AI-Based Drug Discovery in Neurodegeneration

Abstract

Neurodegenerative diseases -- Alzheimer's disease (AD), Parkinson's disease (PD), amyotrophic lateral sclerosis (ALS), and Huntington's disease (HD) -- collectively affect over 50 million people globally and represent the most challenging frontier of drug discovery: all four diseases are characterised by complex, heterogeneous pathophysiology involving protein aggregation, mitochondrial dysfunction, neuroinflammation, and synaptic loss; the blood-brain barrier restricts drug delivery; clinical trials require 5-10 year follow-up for primary endpoints; and the clinical translation rate from preclinical success to approved therapy is < 1% -- the lowest of any therapeutic area. AI-based drug discovery approaches offer specific advantages for neurodegeneration: graph neural networks can predict CNS target druggability from protein structure; multi-omics AI integration can identify disease-stage-specific therapeutic targets from patient omics; and in silico brain organoid modelling (extending BPLA paper #369 organoid AI) can simulate neurodegeneration pharmacology. This study systematically evaluated 284 AI-based neurodegenerative disease drug discovery studies (2,840 target-compound-evidence data points; Italian neuroscience and computational pharmacology groups; AD, PD, ALS, and HD; 2018-2025) comparing AI target identification, compound screening, biomarker discovery, and clinical translation rates. An AI Neurodegeneration Drug Discovery Index (ANDDI) integrating mechanistic target validation, multi-omics disease model depth, BBB penetration prediction, and clinical biomarker alignment predicted clinical phase transition with r = +0.84, identifying tau and alpha-synuclein targeting with AI-guided antibody engineering as the highest-ANDDI neurodegeneration drug discovery approach.