Use R to Create and Execute Reproducible CWL Workflows for Genomic Research

Authors: Qian Liu ^[Roswell Park Comprehensive Cancer Center], Another Author^[Roswell Park Comprehensive Cancer Center].
Last modified: July 27, 2023.

Pre-requisites

• Basic familiarity with DNA-seq data variant calling
• Interest of using workflow language

Workshop Participation

The workshop format is a 45 minute session consisting of hands-on demos, exercises and Q&A.

R / Bioconductor packages used

• ReUseData
• RcwlPipelines
• Rcwl

Workshop: Somatic variant calling

For the somatic variant calling, we will need to prepare the following:

• Experiment data
  • In the format of .bam, .bam.bai files
• ReUsable Genomic data
  • reference sequence file (b37 or hg38)
  • Panel of Normals (PON) ref
• Software tool:
  • Here we use Mutect2to Call somatic SNVs and indels via local assembly of haplotypes. ref

We also want to have the data analysis workflow to be reproducible:

1. Software tool properly tracked for version, docker image etc.
2. Data provenance properly tracked for public data resources for:
   • workflow reproducibility
   • later reuse in other similar projects

The first can be solved by workflow languages (e.g., CWL, WDL, snakemake, etc.). There is no similar tools for the 2nd task.

In this workshop, I will demostrate two Bioconductor packages: Rcwl as an R interface for CWL, and RcwlPipelines for >200 pre-built bioinformatics tools and best practice pipelines in R, that are easily usable and highly customizable. I will also introduce a R/Bioconductor package ReUseData for the management of reusable genomic data.

With these tools, we should be able to conduct reproducible data analysis using commonly used bioinformatics tools (including command-line based tools and R/Bioconductor packages) and validated, best practice workflows (based on workflow languages such as CWL) within a unified R programming environment.