GENIE: A Two-Stage Interpretable Deep Learning Framework for Revealing the High-Order Genetic Interaction Network of Alzheimer's Disease

Alzheimer's disease (AD) is a neurodegenerative disorder characterized by progressive cognitive impairment and memory loss. The underlying mechanisms of AD onset and progression remain unclear. To date, no effective method has been able to uncover the complex, high-order single nucleotide polymorphism (SNP)/gene associations that could elucidate the pathogenesis of AD. Existing computational methods, which are either based on single-site detection with limited analytical capacity or multi-site modeling with low computational efficiency, struggle to handle the growing volume of large-scale, high dimensional data and are only capable of identifying low-order associations. To address these challenges, we propose a novel framework for explaining high-order multi-locus AD associations via deep neural networks, termed GENIE. This method accurately and efficiently calculates high-order associations between SNPs and genes, prioritizing potential interaction networks for downstream biomedical research. Our experimental results are supported by existing research and literature. Additionally, statistical validations of the model and its findings confirm the reliability of GENIE as a hypothesis generation tool.