Improving PRS of autoimmune diseases by inferring effect sizes from associations with other diseases
Approved Research ID: 86585
Approval date: June 13th 2022
While other factors like environment still come into play other than genetics, it is clinically relevant to figure out if a person has a high-risk of developing certain types of diseases, for individual well-being as well as health of the population. Polygenic Risk Score (PRS) is a score calculated to just that, but because research cohorts cannot fully be representative of the population, there are errors in values which lead to hesitance of usage in clinical setting. The goal of the research is to remedy that.
With the recent strides in machine learning, especially those that use neural networks, it became possible to gain new insights from data previously unfounded. If such power can be used to glean new insights not from the disease itself, but from the relationship of diseases, in this case autoimmune diseases, we would be able to refine and better the existing PRS. Then, we would be able to refine the effect of genotypes in determining the odds ratio of diseases.
Because the project does not require monitoring of participants, the research can be concluded in a timely fashion. However, despite the efforts, the time it takes to process whole genome sequencing data, as well as training of the models may take time. Given 36 months the project would be able to reach a conclusion.
With the model shows a good performance, we would be one step closer to create an accurate PRS, such that patients or healthy individuals may be able to understand their risk of contracting a disease with a simple genotyping assay.