Functionally-informed Loci Association Test for disease and phenotype
Current genetic assocation test is still unable to precisely predict health outcome from genetic variants, mainly due to the impact of population structure and that current association test takes into account isolated variants. Here we aim to build a new framework of genetic association test based on functional information, which will capture larger proportion of genetic associations and provide higher prediction accuracy. We reason that according to the central dogma, all non-coding genetic variations could impact a phenotype only if it could impact RNA transcription via functional genomic alterations. Thus, we will first integrate the biological impact of many variants into one score via deep learning models, then test the association between this score and the outcome. We will further build predictive model based on these associations and test its performance in UK Biobank participants with different ethnics. This project will last for 36 months. We expect this project to profoundly promote the application of diseases risk prediction via genome-wide association study and aid early and precise intervention, especially for those non-European populations.