Integration of Genetic and Imaging Data for Alzheimer's Disease Diagnosis and Interpretation

Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by complex interactions between genetic risk factors and structural brain changes. Traditional diagnostic approaches that rely on single-modality data, such as imaging or genomics alone, often fall short in both predictive accuracy and biological interpretability. To address these limitations, AlzCLIP, a novel contrastive learning framework that integrates single nucleotide polymorphism (SNP) profiles and MRI-derived imaging features into a unified embedding space is introduced. This joint representation captures disease-relevant interactions between genetic variation and brain structure, enabling both accurate diagnosis and mechanistic insight into AD. AlzCLIP is trained and evaluated on two large-scale cohorts, the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the UK Biobank (UKB), and demonstrated robust diagnostic performance, outperforming state-of-the-art baselines by up to 19%. More importantly, the model yields interpretable outputs through feature importance and interaction analyses, identifying key contributors to AD risk, including rs1135173, rs7575209, and rs66763080, as well as structural markers such as hippocampal volume and precuneus surface area. Notably, AlzCLIP uncovered genotype-specific effects on imaging phenotypes. Specifically, rs11077054 is associated with increased white matter hyperintensity burden and amygdala atrophy, suggesting a potential link between this variant and AD-related structural brain changes. Together, the results highlight AlzCLIP's potential to enhance AD risk prediction and provide biologically grounded insights by integrating multi-modal genomic and imaging data.