RNAseq pipelines - current status

Created: 2020/09/24 15:19:42 Last modified: 2021/08/03 16:28:54

• Aim: Evaluate the current pipelines available for processing RNA-seq data

Table of contents

• RNAseq pipelines - current status
  • Table of contents
  • Pipelines
  • Notes

Pipelines

See RNAseq pipelines for a spreadsheet of pipelines

Notes

My current favorite pipelines:

• nf-core/rnaseq because
  • It has many stars/forks/contributors - lots of people using and developing it
  • Open source, workflow language, package management, cloud computer support, resource management, good documentation
  • Good analysis - several QC steps, trimming, removal of ribosomal RNA, choice of multiple aligners/quantification tools (STAR, salmon, RSEM, HiSAT2, UMI-based deduplication, mark duplicates, coverage files)
• rna_fq2bam
  • GPU accelerated with parabricks
  • Uses STAR, gatk SortSam, gatk MarkDuplicates under the hood
  • Looks like it would be fairly equivalent to part of nf-core/rnaseq (which also uses STAR, picard MarkDuplicates (which is the basis of gatk MarkDuplicates). However, gatk SortSam would be slightly different to the samtools sort command the nf-core/rnaseq pipeline uses, but I can’t imagine different sorting tools produce very different outputs since they all just sort the data
  • We could develop a subworkflow in nf-core/rnaseq to run (part of the pipeline) in a GPU accelerated mode for those that have a parabricks license
• star_fusion
  • GPU accelerated with parabricks
  • Uses STAR-Fusion under the hood
  • A standalone STAR that is used in rna_fq2bam
• rna_pipeline
  • GPU accelerated with parabricks
  • “GATK best practice”
  • This is a variant discovery pipeline, not an transcript count pipeline such as nf-core/rnaseq
  • This pipeline could be used alongside nf-core/rnaseq if we’re interested in variant discovery
• snakemake-workflows/rna-seq-star-deseq2
  • It has many stars/forks/contributors - lots of people using and developing it
  • Open source, workflow language, package management, cloud computer support, resource management, good documentation
  • Good analysis - several QC steps (custom python scripts and multiqc), automatic download of ensemble sequences/annotation files, SRA fastq files, trimming with cutadapt), trimming, alignment with STAR and differential expression
  • Pro compared to nf-core/rnaseq - does a differential expression/pca analysis (with the deseq2 package and the affycoretool plotPCA function) - I don’t see this as a major benefit since we have all the code to run these analyses, and it would be much easier to customise our plots as well as run a differential expression analysis with additional tools (such as limma/voom) with our own code
  • Con compared to nf-core/rnaseq - no multiple option for aligners/quantification tools, doesn’t look like it does the UMI-based deduplication
• epruesse/ymp
  • Not as popular (at the moment) as some of the other major pipelines but still has several people developing it
  • This is a flexible omics pipeline that has a focus on allowing the flexibility for the user to choose (and explore) the tools they want to analyse their data with
  • Based on Snakemake
  • This could be a really good way to go if it works well