All things interesting, relevant or useful.

Gene Set Enrichment Analysis Best Practices

Link to Paper

• Ludwig Geistlinger, Gergely Csaba, Mara Santarelli, Marcel Ramos, Lucas Schiffer, Nitesh Turaga, Charity Law, Sean Davis, Vincent Carey, Martin Morgan, Ralf Zimmer, Levi Waldron, Toward a gold standard for benchmarking gene set enrichment analysis, Briefings in Bioinformatics, Volume 22, Issue 1, January 2021, Pages 545–556, https://doi.org/10.1093/bib/bbz158

Takeaways

• GSEABenchmarkeR R/Bioconductor package ensures reproducible benchmarking of enrichment methods.
• Enrichment methods developed for microarray data can be applied to RNA-seq data with VST.
• Type of null hypothesis tested can impact gene set testing and identification of enriched gene sets.
• Self-contained methods identify gene sets as enriched containing a single differentially expressed gene; ROAST and GSVA are recommended.
• Competitive methods test for excess differential expression in the gene set compared to the background level and tend to rank relevant gene sets systematically higher; ORA and PADOG are recommended.

Outdated Gene Annotation Nature

Link to Paper

• Wadi L, Meyer M, Weiser J, Stein LD, Reimand J. Impact of outdated gene annotations on pathway enrichment analysis. Nat Methods. 2016 Aug 30;13(9):705-6. doi: 10.1038/nmeth.3963. PMID: 27575621; PMCID: PMC7802636.

Takeaways

• The success of pathway analysis depends on the quality of gene annotations.
• The use of outdated resources strongly affects practical genomic analysis and recent literature.
• Many software tools interpret gene lists using functional information that has not been updated for years.
• In a survey of 25 web-based pathway enrichment tools and citations of these tools in over 3,800 publications, most tools were found to be outdated by several years.

Reproducibility Nature Methods

Link to Paper

• Wratten, L., Wilm, A. & Göke, J. Reproducible, scalable, and shareable analysis pipelines with bioinformatics workflow managers. Nat Methods 18, 1161–1168 (2021). https://doi.org/10.1038/s41592-021-01254-9

Takeaways

• High-throughput technologies have generated massive amounts of biological data, making reproducibility of analysis workflows a key issue in computational biology.
• Workflow managers (e.g. Snakemake, Nextflow, Cromwell) automate data processing and analysis, enabling transparency, code sharing, and long-term reproducibility through data provenance.
• Workflow managers provide detailed information on input parameters, execution environment, software version, resource usage, and pipeline steps, and the workflow itself can be archived and made citable.
• Portability is achieved through package managers (e.g. Conda) and containerization software (e.g. Docker), allowing for platform-independent software installation and distribution.
• Scalability is achieved through efficient resource management and the ability to handle any size and quantity of input data.
• Workflow parallelization and adaptive scheduling are commonly used to manage resources and parallelize workflow steps.