Pathway analysis for proteins without a ranking criteria

Question

[halo22]

Hello All,

I would like to have your opinion on doing pathway analysis when p-values or foldchange or any ranking criteria is missing. Usually when I do my pathway overrepresentation analysis, I try to rank my proteins/genes with fold-change and then filter the proteins with q-values before running pathway analysis, I mostly use the library clusterProfiler(KEGG and Reactome) to do my analysis.

Today I received a list of proteins without pvalues or foldchange values, what would be a good way to perform pathway analysis when the ranking criteria is missing. Please let me know if the question requires more clarification.

Thanks

[ivivek_ngs]

Have you tried EnrichR tool. You just need to feed gene symbols in your case protein names. Obviously with FC or pvalue ranking it's better but it should still work. Even IPA works without any ranking but it's better to put direction or ranking. If you do not have it it's OK . Or you can ask your collaborators if they have any kind of FC or ranking metrics. If not then just try without them. Also I reckon you can feed these list of proteins to String database for finding scores of connectivity from PPI network and co expression profile. You can use that information to rank your proteins in worst case scenario and run standard pathway analysis tools. Hope this is suitable for you.

Answer 1

[h.mon]

Either you need pre-ranked data, or a list of "significant" genes and "background" genes. If your list is pre-ranked, you can use GSEA on any gene set available; if you have "significant" and "background" lists, you can do classic over-representation analysis using Fisher / hypergeometric test.

Is the list you have been given pre-ranked in a meaningful way?

[ivivek_ngs]

h.mon is correct. You need a background list or you need a pre ranked list. So I asked for asking collaborators. Else it will not make much of a sense to be honest. But if nothing is available then you can proceed as I said. However I would not go for traditional pathway analysis rather go for molecular functions with GO analysis without knowing direction or ranking criterion just to have an idea of what functions are enriched with your list. This will give a. Idea to your data but you cannot really bet on them tbh. The PPI network score is another way but just for your list of proteins that are more strongly interacting and you score on them and then do any enrichment. I can only think of this right now but probably more improved method will be suggested by someone. Good luck

[halo22]

I don't have the rank (foldchange or p-val). But, I am told that proteins are statistically significant.

[ivivek_ngs]

They should be significant based on certain criteria. Ask for those informations. Or ask for the background on which they were found to be significant. Even geneSCF is fine

Answer 2

[EagleEye]

Use gene symbols in GeneSCF, Gene Set Clustering based on Functional annotation (GeneSCF)