search

MarioniLab / scran

Clone of the Bioconductor repository for the scran package.
https://bioconductor.org/packages/devel/bioc/html/scran.html
39 stars 23 forks source link

findmarkers summary.logFC values #89

Closed cstrlln closed 4 months ago

cstrlln commented 3 years ago

Hello, noticed that the summary.logFC is just the smallest value among the reported logFC's when using findMarkers. is this as expected? Thank you

LTLA commented 3 years ago

If you read ?findMarkers, you will be redirected to ?combineMarkers, which says (assuming you're using the default pval.type="any"):

For each gene and cluster, the summary effect size is defined as
the effect size from the pairwise comparison with the lowest
p-value. The combined p-value is computed by applying Simes'
method to all p-values. Neither of these values are directly used
for ranking and are only reported for the sake of the user.

If you don't like that definition, you can try some of the other pval.type=, or compute your own statistics and supply them to findMarkers() via row.data. The latter is effectively how summaryMarkerStats() works.