Developing machine learning tools for robust biomarker discovery
Approved Research ID: 48319
Approval date: August 27th 2020
Complex traits, such as BMI, or height, or diseases, such as cardiovascular diseases or cancer, are caused by a combination of genetic predisposition and lifestyle. In the last decades, researchers have identified mutations in many genes that underly many complex traits. For example, we know mutations that make it possible, to a certain extent, to predict a person's height, or increase a person's risk to develop breast cancer. This helps understanding the biological processes involded in the trait of interest, and ultimately, preventing and curing diseases.
However, despite continued efforts, most genetic causes for complex traits remain unknown. One possible explanation is that these complex traits are caused by the combination of subtle mutations in several different genes. These alterations are hard to find individually. However, they are likely to affect the same known cellular mechanism(s), and introducing information about known biological relationships between genes in the study can help.
Another issue is that, to make computations feasible, we generally assume that there is a linear dependency between each of the mutations and the phenotype. But that need not be true.
We are developing statistical tools that model these two aspects, so as to propose algorithms that will help us better understand how complex traits are governed and how diseases appear.
The project will last for 3 years.