This project encompasses two aims tackling key roadblocks impeding the translation of findings from genome-wide association studies (GWAS) into future and personalized medicine.
AIM 1. Accelerate deployment of genomic risk prediction into the clinic
Genomic risk prediction is classically implemented using polygenic scores (PGS), which are linear combinations of hundreds to millions of genetic variants across the genome. The key conditions for deploying PGS in the clinic are: high accuracy, reliability and equity. To ensure these conditions are met, the proposed project aims to progress the development of the most accurate PGS based on whole-genome sequences (WGS) for all common diseases diagnosed in the UK Biobank, and the development of transferable PGS across populations with different ancestries.
AIM 2. Build high-resolution maps of disease-causing genetic variation
Progressing from disease-associated to Disease-Causing Genetic Variation (DCGV) is extremely hard using existing methods and data. Some of the problems at the core of this challenge are the lack of genetic diversity in GWAS, the lack of mathematical understanding of existing methods, and a narrow view of disease classification. This project aims to address these fundamental problems and progress mapping of DCGV across multiple diseases using novel machine learning methods and data from all (sub)populations in the UK Biobank. These methods will leverage the richness of WGS and other omic data available in the UK Biobank in combination with a wide range of phenotypes. To maximize power, our research will develop new statistical methods to analyze multiple phenotypes simultaneously.