GSEA algorithm implementation

[Ferenc Kagan]

Dear GSEA Help Community,

My name is Ferenc Kagan, and I am a metabolomics researcher at Medipredict. We are currently working on developing a functional enrichment analysis in-house for metabolomics datasets and are interested in implementing the GSEA algorithm. However, after reading the original publication (PMID: 16199517), we still have a few questions regarding the algorithm's implementation and would love to hear feedback from this community in regard to those questions.

Could anyone kindly provide us with clarification on the following issues: firstly, we would like to know what criteria the background set is sorted based on. Secondly, we would like to know if there is a need to sort again at the start of each permutation iteration. Lastly, we would appreciate an explanation of the split that occurs between positive and negative correlations, and whether this split is necessary in the first place.

We would also like to inquire whether the random-walk should start from 0 and end at 0.

Thank you in advance!

Best regards,

Ferenc Kagan