A weird result: Highly enrichement visually but p value is not significant
chent5 opened this issue · 6 comments
Hi,
Thanks a lot for developing this package. This package is very helpful and easy to use. Recently, I got a weird result as below figure:
Visually, the turquoise geneset is highly enriched in the negative side of the ranked genes. However, p-value = 0.16, which doesn't make sense to me.
I tried ks test, the statistics are D = 0.8891, p-value < 2.2e-16. The p-value of the KS test is extremely significant.
Could you help to solve this issue?
Thank you so much!
It does look strange indeed. Can you upload the inputs (gene set and statistics vector) here, so that I can reproduce the behavior?
Thanks a lot for your quick response. Do you have a preferred email to receive the data?
You can send the data to alsergbox@gmail.com
I have sent the data to your email. Thank you so much for your help in advance!
Hi, sorry for the delay. I've checked your data, your ranks
vector is highly unbalanced and skewed towards positive values, that's causing some troubles. There is an appropriate warning message if you use fgseaMultilevel
and not the fgseaSimple
version.
set.seed(123)
fgseaRes <- fgseaMultilevel(pathways = geneset,
stats = rank,
minSize=15,
# maxSize=500,
nPermSimple=100000
)
The code above gives the following table:
pathway pval padj log2err ES NES size
1: turquoise 1e-50 1e-50 NA -0.9566721 -4.031049 857
Thanks so much for your help!