Sequencing

11th Annual Sequencing, Finishing, and Analysis in the Future Meeting
SCALABLE AND EXTENSIBLE NEXT-GENERATION
SEQUENCE ANALYSIS PIPELINE MANAGEMENT FOR
OVER 50,000 WHOLE-GENOME SAMPLES
Friday, 3rd June 17:00 La Fonda Ballroom Talk (OS‐10.04)
Jesse Farek, Adam English, Daniel Hughes, William Salerno,
Kimberly Walker, Donna Muzny, Richard Gibbs
1 Human Genome Sequencing Center Baylor College of Medicine
The Baylor College of Medicine Human Genome Sequencing Center (HGSC) has recently added
Illumina HiSeq X Ten sequencers to its sequencing fleet, which currently processes over 2,000 wholegenome
samples per month. These samples originate from multiple projects and collaborators, including
the Alzheimer’s Disease Sequencing Project, the Trans‐Omics for Precision Medicine Program,
Baylor Miraca Genetics Laboratory, the Centers for Common Disease Genomics, the CHARGE
Consortium, and the Center for Mendelian Genomics. In order for sequence analysis at the HGSC
to scale to this increased workload, numerous improvements have been made to the efficiency and
reliability of the center’s sequence analysis infrastructure. HgV is the workflow management system
for primary and secondary Illumina sequence analysis at the HGSC and features tiered XML
pipeline protocols, job tracking, LIMS communication, verbose logging and stable reproducibility.
HgV’s protocol definition and LIMS communication infrastructure has been reworked for greater
configurability so that pipeline protocol and LIMS parameters can be easily modified to accommodate
different project requirements. Specifically, the HGSC has configured HgV use both local and
cloud‐based compute resources and to enforce CAP‐ and CLIA‐compliant data handling for clinical
pipelines. Secondary analysis programs have been rewritten for increased computational efficiency.
The Atlas2 SNP and Indel variant callers (originally written in Ruby) have been rewritten and combined
into a single C++ program that runs on average more than 50 times faster than Atlas2 and
with improved variant calling quality and consistency. Two new custom reporting programs, SeqAnalyzer,
which calculates FASTQ sequence metrics, and AlignStats, which calculates BAM alignment
and coverage metrics, have been written to use significantly fewer computational resources than the
existing programs they replace. Other areas of improvements to analysis workflows have also been
investigated, including measuring the effects of local Indel realignment and base quality recalibration
on variant call quality and researching efficient N+1 joint calling solutions for creating project
level VCFs. These improvements have resulted in fast, extensible, and easily manageable analysis
pipelines for human resequencing and other applications on the HiSeq X platform that have allowed
the HGSC to concurrently support the heterogeneous analysis requirements of multiple large‐scale
sequencing projects. To date, HgV has managed the analysis of over 5,000 whole‐genome samples
and is expected to handle over 50,000 more samples in the near future.
152