Which function is best for pathway analysis?

Question

[Chris]

Hi Biostars,

I found there are many function to perform pathway analysis such as fgsea(), gseGO(), gseKEGG(), enrichGO() which made me quite confuse which result I should focus on. Should we focus on pathways that share between each function? Getting a correct background gene set is important. However, how can we find the background gene set for our experiment. I download the background gene at https://www.gsea-msigdb.org/gsea/msigdb/human/collections.jsp#C3 Only fgsea() mention about background gene but not customize for each experiment. I appreciate your help!

Answer 1

[Yogi]

Few things to mention.

Pathway Enrichment Analysis can be split into two categories.

1. Functional Class Scoring Methods
2. Overrepresentation Analysis

"GSEA" is #2 above whereas some of the others above are #1.

Regarding background data set, I'll give an example for you to consider.

If you do differential RNA-Seq analysis, you might only take the genes that are "actually expressed". This could be all genes where the average normalized expression across all samples is above some threshold "x".

[Chris]

Thank you for your reply! Do you have code that I and others can implement? I am not sure how to customize the gmt file from my differential expressed genes. From the functions I used, I see only enrichGO() used a cut off so I think only enrichGO() is in #2.

[luffy]

Chris Please look into @Hamid Ghaedi's well documented post on ORA and GSEA, Or this documentation. Hope this helps.

[Chris]

Thanks luffy! Great material but seem they didn't mention how to customize the background genes depend on our data.