A Novel Deep Learning Framework for Alzheimer’s Gene Prediction via SNP and Gene Network Integration

Alzheimer’s disease (AD) is a complex neurodegenerative disorder whose genetic basis remains incompletely understood. Early and precise diagnosis is crucial for effective intervention and improved treatment outcomes. Conventional diagnostic techniques, including neuroimaging and cerebrospinal fluid analysis, are often invasive, costly, and slow. In contrast, gene network-based approaches have emerged as promising tools for predicting AD-associated genes. However, existing methods struggle to fully capture the intricate relationships between genetic variations and disease pathology. To overcome these limitations, we introduce Gene-SNP-Net-Model. This advanced deep learning framework synergistically integrates single-nucleotide polymorphism (SNP) data with functional gene networks using graph convolutional networks (GCNs). Our model processes SNP data through specialized neural network layers while leveraging GCNs to encode gene-gene interactions. These features are fused into a unified representation, followed by a classification layer that predicts gene-disease associations. Extensive evaluation demonstrates that Gene-SNP-Net-Model surpasses current state-of-the-art methods, achieving an AUROC of 0.9737 and 98.20% accuracy. These results underscore its ability to decipher both linear and nonlinear genetic interactions, enabling highly accurate prediction of AD risk genes. Furthermore, molecular validation confirms that the top-ranked genes identified by our model exhibit strong associations with AD, offering novel insights into its genetic architecture. By unifying SNP data with gene networks, Gene-SNP-Net-Model provides a powerful tool for early AD detection and paves the way for precision medicine strategies in neurodegeneration.