Case study
Turning genomic risk results into clinical information
Data-pipeline infrastructure supporting an NIH-funded study of genetic risk across ten chronic medical conditions.
For genomic information to become useful in clinical settings, scientific results have to move through reliable operational systems. During my time within the Broad Institute's Genomics Platform, I developed data-pipeline infrastructure supporting an NIH-funded study of genetic risk across ten chronic medical conditions.
01 · The broader program
A scientific program with an operational layer.
eMERGE studied the clinical implementation of genomic information across a national research network. My role was within the software and data infrastructure supporting that work—not the design of the scientific study or the invention of its polygenic risk scores.
02 · The engineering problem
Turn scientific outputs into a repeatable workflow.
Results had to move through processing and quality controls before they could support clinical reporting. Reliability at each handoff was part of making the information useful.
- Genetic risk results
- Automated processing
- Quality checks
- Clinical reporting
- 10 clinical sites
- 25K+ participants
03 · My contribution
Software and data infrastructure
I built infrastructure within the data pipeline for result processing, quality checks, and support for clinical reporting. The broader program included more than 155,000 genetic risk results, 25,000 participants, and 10 clinical sites.
The distinction matters: the scientific and clinical program defined what information was meaningful; the infrastructure helped that information move through a dependable operational path.
04 · The system around the science
Engineering inside a scientific program
05 · Publication
From infrastructure to published research
I co-authored the resulting Nature Medicine publication, reflecting the software and data infrastructure contribution within a much larger scientific and clinical collaboration.
Nature MedicineSelection, optimization and validation of ten chronic disease polygenic risk scores for clinical implementation in diverse US populationsRead the publication ↗ (opens in a new tab)What I learned
Correct results still need a reliable path to use.
In healthcare, a technically correct model or result is only one layer of the product. Information also has to move through dependable systems, quality controls, and clinical workflows before it becomes useful.