When we run the same compound lists through both the functional and pathway analysis modules, the enriched-pathway outputs differ in both composition and p-values. In the functional analysis, all pathways have p-values above 0.05, whereas in the pathway analysis they fall below 0.05. Could you help me understand why these results diverge?
Are you referring to Functional Analysis (untargeted) vs. Pathway Analysis (targeted)? This is interesting as people do either way but not both in general. Keep in mind that these two modules use different algorithms (mummichog for untargeted) and over-representation analysis (ORA for targeted) - their p-values should be different. However, I hope the rank of the pathways will remain similar. My gut feeling is that mummichog uses permutation (non/semi-parametric) which is usually less sensitive compared simple ORA (parametric)
If you are referring to Enrichment Analysis vs. Pathway Analysis (both targeted). Please make sure that they are using same library. Note the KEGG pathway library was recently updated in Pathway Analysis module.