search

BostonGene / MFP

Mollecular Functional Portraits
Other
33 stars 13 forks source link

ssgsea score #2

Open ashley0207 opened 3 years ago

ashley0207 commented 3 years ago

Could you please provide why you calculate your ssgsea score in this formula "(sranks 1.25).sum() / (sranks 0.25).sum() - (len(ranks.index) - len(common_genes) + 1) / 2"? Thank you.

avkitex commented 2 years ago

Hi, it's just simplified formula from the original publication. You may check, It produces exactly the same values as the original ssgsea from the gsva R package

LPH-BIG commented 2 years ago

I have check, but I got the different results from your ssgsea code and gsva R package when using the same input. Could you please give me some ideas?

avkitex commented 2 years ago

I have check, but I got the different results from your ssgsea code and gsva R package when using the same input. Could you please give me some ideas?

Hi, sorry for the late response Is the difference big? Could you please post some images with per-sample comparison?

BTW there could be a little deviation that is due to different ranking in case of equal values (in case of RNAseq = zeros). The difference is little in case of "good" signatures. Usually good signatures are >=5 genes with relatively high expression which correlate with each other (ssGSEA asumption)