The Monthly Researcher Spotlight is our section highlighting the exciting work of the UK Biobank Research Analysis Platform user community. If you would like to be featured, email ukbiobankrap@dnanexus.com.
This was simultaneously published in the April 2024 UK Biobank RAP Newsletter. You can sign up for future installments here.
This month's Spotlight features the Scientific Insight Engineering Group at Genentech and they collaborate with scientists to power scientific discovery and accelerate decision making using computational methods.
(Left) Xiucheng Quek
Bioinformatics Cloud Software Engineer
Genentech, Inc
(Right) Sai Reddy Achakkagari
Bioinformatician
Genentech, Inc
What are the focus and discovery highlights of your research?
We are a research group that collaborates closely with scientists, employing computational techniques to extract scientific insights from their experiments. We develop, manage and execute large-scale computational workflows and methods in a reproducible framework. Our primary workflow development language is the Workflow Description Language (WDL), and we have constructed a framework enabling highly reproducible development, testing, and execution of our workflows. Our work has helped drive various R&D efforts and drug development in the company.
What are some of the key questions that you are looking to answer using UK Biobank data?
The UK Biobank cohort provides a rich set of phenotypes and has whole exome sequencing data for about 470,000 patients. We are interested in correlating certain known variants to specific phenotypes that are available in the UK Biobank cohort. This will provide our scientists with insights into disease biology and might open up pathways to new treatments.
How has the UK Biobank Research Analysis Platform (UKB-RAP) helped you perform your research?
The UK Biobank Research Analysis Platform (UKB-RAP) allowed us to execute workflows at scale.
We started our development at the end of December. By January, following several rounds of refinement and optimization, we initiated the processing of the entire cohort and successfully completed our runs before the month's end. At peak performance, we were processing 1000 jobs at a time, utilizing up to 8 cpu per job.
Because the UKB-RAP platform can run Workflow Description Language (WDL) workflows, we successfully transferred our internally developed and thoroughly tested workflow to the platform. For further optimization, we employed the smart-reuse feature during the pilot stage. This enabled us to iterate various workflow versions efficiently by re-running only the components that had changed between different versions. Additionally, we used dxFuse to mount our reference data, effectively reducing the time needed for data localization and thus, computation time.
