A technical contribution at Illumina as part of a team effort that resulted in a published patent application for a novel machine learning pipeline.
The Problem#
Genome-wide association studies (GWAS) are powerful for detecting variants associated with disease risk. However, transitioning from “associated variants” to identifying the actual causal genes is a notoriously difficult bottleneck in precision medicine.
The Solution#
The team developed a machine learning framework that predicts causal genes directly from GWAS summary statistics. By utilizing advanced ML techniques, this approach substantially improves gene identification performance in terms of both precision and recall compared to traditional methods.
Key Contributions#
- Pipeline Architecture: Designed and implemented the reproducible ML pipelines on HPC and AWS infrastructure, including the Hail/Python code for the GWAS data processing component.
- Scalability: Contributed to optimizing the system to handle the massive datasets characteristic of modern genomics.
- Innovation: Co-authored the underlying methodology described in the patent application.
Patent Details:
- Title: Machine learning pipeline for genome-wide association studies
- Publication No: US20240120024A1
- Assignee: Illumina Inc.