Network Medicine in Complex Disorders

Abstract

Complex disorders -- type 2 diabetes, Alzheimer's disease, inflammatory bowel disease, and major depressive disorder - arise from the interplay of hundreds to thousands of genetic variants, environmental exposures, and molecular pathway perturbations that cannot be understood through single-gene or single-pathway analysis. Network medicine maps these disorders onto the human interactome -- the complete network of molecular interactions within cells -- revealing that disease-associated genes cluster into localised network neighbourhoods (disease modules) whose topology predicts comorbidity patterns, drug targets, and patient subtypes. We present the Network Medicine Assessment Framework (NMAF), evaluating five network-based analytical approaches -- protein-protein interaction (PPI) network propagation, gene regulatory network inference, multi-layer network integration, graph neural network disease module detection, and causal network inference from multi-omics data -- across four complex disorders (type 2 diabetes, Alzheimer's disease, Crohn's disease, and major depressive disorder). Our Network Medicine Effectiveness Score (NMES) measures disease module identification accuracy, drug target prioritisation, patient stratification capability, mechanistic interpretability, and experimental validation rate. Multi-layer network integration achieves the highest NMES (0.924) by combining PPI, transcriptomic co-expression, metabolomic correlation, and genetic association networks into a unified framework, while causal network inference achieves the highest mechanistic interpretability (0.960) through identification of directional regulatory relationships that distinguish disease drivers from downstream consequences.