Alex Diaz-PapkovichApr 17, 2023
SFGravel
Shadi Zabad’s latest work on predicting genetic risk:
https://www.cell.com/ajhg/fulltext/S0002-9297(23)00093-9
Variational Inference for Polygenic Risk Score (VIPRS) consistently competes with or outperforms leading approaches.
Fast and accurate Bayesian polygenic risk modeling with variational inference

We present VIPRS, a fast and accurate variational Bayesian method for estimating polygenic risk scores from genome-wide association study (GWAS) data. The method is shown to be robust and competitively accurate against popular baselines and scales well to dense genotype array data.

The American Journal of Human Genetics
Apr 12, 2023 at 9:23pmWeb
Show threadSFGravelApr 12, 2023
Because the approach is very scalable, Shadi was able to test the method in a wide range of scenarios and traits – real and simulated, matching and different ancestry, in and out of cohort, and for discrete and continuous traits. It performs well across the board. The best improvement was for LDL, where VIPRS it outperformed other tested methods because it correctly identified large effect variants.
Show threadSFGravelApr 12, 2023
I thought that the most important determinant of PRS performance would be high-level model choices. Turns out that a lot of “small” decisions on how we optimize parameters, encode LD, and filter data have a large effect on performance. Shadi worked hard to design robust code.
Please check it out!
https://github.com/shz9/viprs
https://gravellab.github.io/publications/
GitHub - shz9/viprs: Variational Inference of Polygenic Risk Scores

Variational Inference of Polygenic Risk Scores. Contribute to shz9/viprs development by creating an account on GitHub.

GitHub