pantherdb / Helpdesk

0 stars 0 forks source link

Redoing a Panther analysis with older version #38

Open Echodonut opened 1 year ago

Echodonut commented 1 year ago

Hi!

I am trying to redo some panther analyses, to compare with ones I did last year. To this end I am trying to use the older version of Panther that was out at the time, v17, through https://archive.pantherdb.usc.edu/

However, I cannot get the same results despite using the same set of genes. Here are the results from a set of analyses, with the same geneset, for one of the GO terms that is reported for the Panther overrepresentation test, biological processes (same test types and corrections of course):

GO term: biological regulation (values are: # known genes for the GO term; # genes in my set; # expected; Fold enrichment; over or underrepresented; raw P value; adjusted P value)

  1. GO analysis (performed again as control, on 12-10-2023) 4349 790 677.99 1.17 + 9.39E-06 4.90E-04

  2. panther 18, updated 17-09-2023 4349 790 677.99 1.17 + 9.39E-06 4.90E-04

  3. Panther 17, updated 23-02-2022 4378 798 683.74 1.17 + 6.52E-06 3.43E-04

  4. My older analysis from 22-04-2022 4448 811 694.67 + 1.17 5.28E-06 2.94E-04

I expected that the analysis with the archived version of Panther 17 should give me the exact same output as the analysis which I did on 22-04-2022 I would like to ask for some help to figure out what is going on.

mugitty commented 1 year ago

@Echodonut, If 'GO biological process' was used as the annotation data set, these are updated ~monthly. Therefore, it is possible for the results to change.

Echodonut commented 1 year ago

@mugitty Thanks! That is useful to know, it pinpoints where I should be directing my effort.

Would you know a way I could use the specific older version of GO & PANTHER? I already asked at the GO Github forum, and was forwarded here for potential advice.

mugitty commented 1 year ago

@Echodonut the PANTHER GO-SLIM annotation data sets are more stable. They change ~yearly when a new version of the PANTHER library is released

Echodonut commented 1 year ago

@mugitty I will do this for the next publication. For the one I am currently working on, I might just have to either redo them, or just accept that the values are not going to be reproducible and use what I have logged. Either way is probably fine.

Cheers,