H3: GSEA

Objective

Perform a GSEA a ranked gene list of mesenchymal and immuno-reactive ovarian cancer subtypes

Started: 2023-04-07 | Completed: 2023-04-07
Estimated time: 2 hrs | Actual time: 1.5 hrs

Procedure

• Software Download
  • I was not able to run this on Docker so I had to download the software (GSEA v4.3.2 Mac App) on the GSEA official website, along with Cytoscape and the EnrichmentMap pipeline collection.
• Data Download
  • Gene list is available at https://github.com/bcb420-2020/Student_Wiki/blob/master/MesenvsImmuno_RNASeq_ranks.rnk.
  • Annotation data is available from the Bader Lab geneset collections. Human_GOBP_AllPathways_no_GO_iea_March_02_2023_symbol.gmt file.
• Run GSEAPreranked and set the parameters:
  • rnk : the ranked gene list
  • No Collapse
  • gmt: the Annotation file
  • Number of permutations: 500
• Parameters for Basic fields:
  • Enrichment: Weighted
  • Maximum size: 200
  • Minimum size: 15

Results

1. Explain the reasons for using each of the above parameters.

• We wanted to use the newest gene set from the Bader Lab as it is updated monthly. It includes GO Biological Process terms for more credible pathways rather than electronic annotations. Therefore, we can use a set of high quality gene sets to use for analysis.
• We set the maximum gene set size to 200 to exclude very large pathways and obtain specific pathways. We set the minimum gene set size to 15 to ensure that the gene sets are not too small.
• We set collapse/remap to gene symbols as No_Collapse as they have already been matched to symbols. We do not want to remap them.
• We chose gene set permutation as we used a pre-ranked gene list and we do not want to lose this ranking of the genes after GSEA permutation.

2. What is the top gene set returned for the Mesenchymal sub type? What is the top gene set returned for the Immunoreactive subtype?

• The top gene set for the Mesenchymal subtype is HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION%MSIGDBHALLMARK%HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION.
  • p-value: 0.000
  • ES: 0.87
  • NES: 2.59
  • FDR: 0.000
  • 147 genes in the leading edge.
  • Top gene associated with this gene set is FBN1.
• The top gene set for the Immuno-reactive subtype is HALLMARK_INTERFERON_ALPHA_RESPONSE%MSIGDBHALLMARK%HALLMARK_INTERFERON_ALPHA_RESPONSE.
  • p-value: 0.000
  • ES: -0.86
  • NES: -2.91
  • FDR: 0.000
  • 79 genes in the leading edge.
  • Top gene associated with this gene set is PROCR.

Conclusion

• GSEA is a non-threshold gene set analysis – it does not exclude genes based on arbitrary cutoffs and instead allows for uncovering minor pathways shadowed by larger pathways.
• It is, however, very important to set an appropriate threshold as it can significantly affect results.
• GSEA results in a list of biological pathways but these terms might still be very generic and not informative.
• I liked the GUI as there is a lot of useful information and is user friendly. The results are also straight-forward and easy to interpret.

References

Mootha, V. K., Lindgren, C. M., Eriksson, K. F., Subramanian, A., Sihag, S., Lehar, J., Puigserver, P., Carlsson, E., Ridderstråle, M., Laurila, E., Houstis, N., Daly, M. J., Patterson, N., Mesirov, J. P., Golub, T. R., Tamayo, P., Spiegelman, B., Lander, E. S., Hirschhorn, J. N., Altshuler, D., … Groop, L. C. (2003). PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nature genetics, 34(3), 267–273. https://doi.org/10.1038/ng1180