Deep learning based causal inference methods for variant interpretation
Approved Research ID: 79957
Approval date: March 10th 2022
A generic question in genetics is whether a non-coding variant affects a particular disease. Genome-wide association study (GWAS) is an approach used in genetics research to associate specific genetic variations with particular diseases. However, correlation does not imply causation.Therefore, it is hard to interpret the GWAS results. Thanks for the recent breakthroughs in combination of deep learning and statistics, we try to examine the effects of the genetic variants that are associated with the biomarker on the disease outcomes. If the individuals with the genetic variants associated with specific biomarkers are observed to have increased disease risk, one could infer that the biomarker is causally linked to the disease. Here, genetic variants are ideal choices of instrumental variables because they are inherited from parents and are not subject to the influences of confounding variables.
Inferring causal relationships is a more significant theoretical and computational challenge than measuring correlations, so our model will combine casual inference theory and deep learning technique. We expect to produce more accurate, robust, and equitable tools for genetic prediction of disease risk. As a basic methodological tool, it will enable individual level disease risk estimation and early interventions for complex diseases such as cancer, hypertension, and cardiovascular disease.