Multi-Disease Deep Learning Framework for GWAS: Beyond Feature Selection Constraints

TL;DR

This paper introduces a scalable deep learning framework for multi-disease GWAS that avoids feature selection constraints and data leakage, effectively modeling complex genetic interactions across multiple diseases.

Contribution

It presents a novel multi-label deep learning architecture for GWAS that outperforms existing methods under strict no-leakage conditions and leverages shared genetic architecture.

Findings

Achieved AUC 0.68-0.96 on five diseases with 5 million SNPs and 37,000 samples.

Demonstrated that architectural choices can outperform existing GWAS methods without data leakage.

Extended the approach to jointly model multiple diseases, improving discovery and efficiency.

Abstract

Traditional GWAS has advanced our understanding of complex diseases but often misses nonlinear genetic interactions. Deep learning offers new opportunities to capture complex genomic patterns, yet existing methods mostly depend on feature selection strategies that either constrain analysis to known pathways or risk data leakage when applied across the full dataset. Further, covariates can inflate predictive performance without reflecting true genetic signals. We explore different deep learning architecture choices for GWAS and demonstrate that careful architectural choices can outperform existing methods under strict no-leakage conditions. Building on this, we extend our approach to a multi-label framework that jointly models five diseases, leveraging shared genetic architecture for improved efficiency and discovery. Applied to five million SNPs across 37,000 samples, our method achieves competitive predictive performance (AUC 0.68-0.96), offering a scalable, leakage-free, and biologically meaningful approach for multi-disease GWAS analysis.