Would different enrichment tools result differently by defining pathway differently

Question

[CY]

I am a bit confused by these two methods.

By definition, gene set enrichment analysis (GSEA) uses a priori gene sets that have been grouped together by their involvement in the same biological pathway. This seems the same as how pathway analysis choose the predefined gene list (by pathway). Can anyone eliberate any differences between this two methods.

I know that GSEA use 'signature' to represent specific biological process while other resources like panther or GO enrichment analysis may uses different gene set to represent the same biological process. Would this cause conflict? Different tools may get different results from the same input gene set.

Answer 1

[Jean-Karim Heriche]

Pathway enrichment or pathway analysis is the approach of interpreting large gene lists in terms of gene functional annotations. Gene set enrichment analysis (abbreviated GSEA) is just one of many methods for doing this.

[CY]

So GSEA is one of the pathway analysis method just like KEGG, Panther or GO Enrichment Analysis. The further question is that if all these methods are here for pathway analysis and each of these sources define each pathway differently (uses different gene set for same pathway). Would this cause conflict? I will get different results using the same input gene set.

[Jean-Karim Heriche]

GSEA is a method whereas KEGG, Panther and GO are databases sources of annotations. GSEA can be used with any source of annotations because essentially, it tests if a predefined set of genes behaves differently between two experimental contexts. You could select the query set of genes in GSEA to represent any pathway from any database. Different methods used on the same source of annotation will generally give different results and using the same method with different sources of annotations will also give different results. Indeed, different sources have different notions of what a pathway is. There are two main sources of differences: frequency and level of curation (e.g. whether the annotation is up-to-date) and granularity (e.g. where one resource will have a pathway named DNA replication with many genes, another resource would cover it with several pathways such as initiation of DNA replication, DNA elongation...).

[CY]

Very informative clarification. Thx :)