Common genetic variants, such as SNPs and INDELs, are major drivers of phenotypic diversity among individuals. Although GWAS have identified numerous associations between these variants and complex traits, most are non-coding, and their functional impact remains unclear. The “missing heritability” phenomenon highlights the limitations of current approaches in fully explaining the genetic basis of complex diseases.
Recognizing the importance of coding variants in shaping protein function and disease susceptibility, this study aims to utilize advanced methodologies to explore their contribution to disease risk and gain a deeper understanding of the genetic architecture of complex traits.
We hypothesize that analyzing common coding variants with advanced machine learning methods will improve our understanding of genetic predisposition and enhance risk prediction for a wider range of complex diseases.
Specifically, we will integrate EWAS approaches with cutting-edge technologies like Protein Language Models (PLMs) and Biologically Informed Neural Networks (BINNs) to analyze germline variant data. By combining these approaches, we aim to extract deeper insights, identify novel associations, and improve risk prediction.
This study will use genotype data from whole-exome sequencing in the UK Biobank. We will assess the performance of our models in predicting disease risk across diverse disease categories and compare their performance with traditional GWAS.
If successful, this study could uncover new genetic associations between coding variants and complex diseases, enhancing our understanding of their genetic architecture and potentially addressing part of the missing heritability. This could ultimately improve risk prediction models and support the development of better early diagnosis and prevention strategies for a wider range of diseases.
Results will be shared with the scientific community through oral presentations and peer-reviewed publications.
