在 F2 实例上使用 Nextflow 和 AWS Batch 构建 Illumina DRAGEN 流水线.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.C M P 3 5 3Marissa E Powers PhDIllumina DRAGEN pipelines on F2 instances with Nextflow&AWS BatchSh

2、e/herSr.Solutions Architect,HPC life sciencesAWSSean ODellCentre for Genomics ResearchDiscovery Sciences,R&D,AstraZenecaCambridge,UK 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Today Why we built this What we discovered How we built it(demo)2025,Amazon Web Services,Inc.or its

3、affiliates.All rights reserved.But first:Who!With much-appreciated help from:Heejon Jo(Illumina)Shyama Mehtali(Illumina)Natalia Jimenez(AWS)Eric Allen(AWS)Omar Khan(AWS)Gabriel Hernandez(AstraZeneca)Marissa E Powers PhDSolutions Architect,HPC life sciencesAWSBoston,MAManu PillaiSolutions Architect,H

4、PC life sciencesAWSCambridge,UKSean ODellCentre for Genomics ResearchDiscovery Sciences,R&D,AstraZenecaCambridge,UK 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Petascale Genomics at AstraZenecaHighly optimised

5、for storing PBs of OMICs dataData lifecycle protocols leverage range of storage classes available within AWS25 PB compressedOptimised Iceberg on S3 variant data store for analysis2.5 trillion rows(and growing)Industry leading genomics processing pipelineObserved peak performance:38,000 WES/day11,000

6、 WGS/dayWES samples:867,188WGS samples:780,5141,647,702 DRAGEN jobs on F1Standardised data pipelines for WES,WGS,RNA,ProteomicsWorkflow orchestrator,file store,data lake,tertiary analysis,deployed and tested with code pipeline.Region-distributed processing to maximise use of DRAGEN on AWS FPGA insta