What's the issue with FDR ? about hyper HOT 10 CLOSED

All of the geneset symbols are upper case, so convert your signature genes to upper case as well.

from hyper.

I am still getting the similar result with "hypergeometric" test ,all the pathways are coming with pval and fdr value of 1 and not showing any gene overlap, however its working with "kstest" .

from hyper.

If your symbols are upper case, and the p-values are 1, then your signature may not be enriched with any genesets.

from hyper.

You could try another geneset compendium, like Hallmark or Reactome.

from hyper.

I think there is some issue, when I am using "hypergeometric", else everything is working fine with "kstest"

from hyper.

Could you send me a script to reproduce your issue?

Or reply with your code and show what the input signature looks like for both the hypergeometric and kstest methods. That way I can let you know if it is your data or an issue with the package.

from hyper.

DE_genes_ranks.txt

library(hypeR)
ranks  <- deframe(read.table("DE_genes_ranks.txt",sep = "\t",header = T))
head(ranks)
msigdb_path <- msigdb_download_all(species="Mus musculus")
KEGG <- msigdb_fetch(msigdb_path, "C2.CP.KEGG")
hyp_obj <- hypeR(ranks, KEGG, test="hypergeometric")
hyp_show(hyp_obj)
hyp_obj <- hypeR(ranks, KEGG, test="kstest")
hyp_show(hyp_obj)
Reactome <- msigdb_fetch(msigdb_path, "C2.CP.REACTOME")
hyp_obj <- hypeR(ranks, Reactome, test="hypergeometric")
hyp_show(hyp_obj)
hyp_obj <- hypeR(ranks, Reactome, test="kstest")
hyp_show(hyp_obj)

from hyper.

This is working now on my end, you need to convert your rankings to a character vector of gene symbols when using test="hypergeometric". See my comments.

library(tidyverse)
library(hypeR)

ranks  <- deframe(read.table("DE_genes_ranks.txt",sep = "\t",header = T))
head(ranks)

msigdb_path <- msigdb_download_all(species="Mus musculus")
KEGG <- msigdb_fetch(msigdb_path, "C2.CP.KEGG")
hyp_obj <- hypeR(ranks, KEGG, test="kstest")
hyp_show(hyp_obj)

# Hypergeometric method takes a character vector of gene symbols
# This should be a differential expression signature for example (top 100 differentially expressed genes)
signature <- names(sort(ranks, decreasing=TRUE)[1:100])
head(signature)

hyp_obj <- hypeR(signature, KEGG, test="hypergeometric")
hyp_show(hyp_obj)

from hyper.

Thanks Federico, now I got it. Hypergeometric test would be overlap of signature genes with genesets. That's why there were no overlap and hence pval were coming as 1.
Nice simplified and comprehensive tool for gene set enrichment analysis with better visualisation.

Thanks
Gyan

from hyper.

Thanks :)

from hyper.