Cloud-Native Variant Calling Pipeline

Challenge

Rebuilding a whole-exome variant calling workflow so it's reproducible and cost-bounded at scale, rather than a one-off script run by whoever last touched it.

Approach

A containerized Nextflow pipeline following GATK best practices, benchmarked on public 1000 Genomes exome data on cloud spot instances.

What we found

The rebuilt pipeline ran faster and cheaper than the pipeline it replaced. The baseline it's compared against isn't documented publicly, which is why that comparison is flagged below rather than stated as fact.

Limits of this demonstration

This is an R&D demonstration on public 1000 Genomes data, not a client engagement. No repo or benchmark script is linked yet, which is why the metric above is flagged TODO-VERIFY. Spot-instance pricing also varies by region and time, which this figure does not account for.