search

jokergoo / simplifyEnrichment

Simplify functional enrichment results
https://jokergoo.github.io/simplifyEnrichment
Other
108 stars 16 forks source link

test with more ontologies #11

Closed jokergoo closed 4 years ago

jokergoo commented 4 years ago

tested:

More to test

selcant commented 3 years ago

Hi! Can we use KEGG/Reactome ontologies with the development version?

jokergoo commented 3 years ago

You can use other ontology databases, but they do not work as well as GO. See results at https://simplifyenrichment.github.io/examples/

The main problems are:

  1. In the simplifyEnrichment paper, it shows the semantic similarity works very well, which is only workable for GO terms, but not for KEGG/Reactome terms. For KEGG/Reactome terms, you can only calculate similarities purely by gene overlap, which normally do not have such "diagonal block" patterns which are used to cluster terms.
  2. for KEGG pathways, I think it not necessary to cluster them. Normally you will not have too many enriched pathways, say, around 20 or 40, which you can already read them one by one. And terms less than 50 are not suggested to cluster.
selcant commented 3 years ago

Thank you! I think I figured it out. Unfortunately I do end up with many KEGG/Reactome terms (>100) in some cases and need to simplify. From what I can see KEGG clustering worked very well, clear diagonal. Trying to get the reactome clustering to work, running into some errors at the moment. Will open a separate issue if I can't make it work. Thanks again for the reply.