Genetic risk prediction for complex diseases by multi-task deep learning model
Aim: Many human diseases are heritable. When our family has disease history, our offsprings may be in risk. How to find this kind of risk factors (so called genetic risk) is an important task. The project aims to produce advanced computational models to predict genetic risk for human diseases.
Complex traits such as diabetes, stroke, cholesterol are polygenetic heritable. Their heritability are contributed by a large proportion of genetic variations. The conventional computational approach has been proven to be effective for predicting genetic risk of diseases (e.g. Haemophilia) associated with single gene or very few genes, but restricted for complex diseases (e.g. diabetes, stroke) associated with many genes. So our work will be to develop a new advanced computational method to predict the genetic risk for these diseases. We also want to discover new risk variants and risk genes associated with the diseases.
This project needs 3 years.
Public health impact:
The research will produce a computational method to predict genetic risk for diseases. The method will be able to identify risk variants and risk genes associated with diseases that may not be identified by previous studies, and provide insight into the development of genetic basis of complex traits.