Hi,

I know that DEseq2 can do low counts filtering but I wanted to set my threshold based on CPM values rather than count values then do filtering. Is there a way in DEseq2?

Thanks for the help.

J.

First you can turn off independent filtering like so:

results(dds, independentFiltering=FALSE)

Then you can choose your own threshold and adjust the p-values yourself:

filterGenes <- res$baseMean < x

res$pvalue[ filterGenes ] <- NA

res$padj <- p.adjust( res$pvalue, method="BH" )

The one caution is that, if the groups are very unbalanced, say 100 vs 10, then you need to be careful not to remove genes which have expression only in the smaller group. These could be truly DE, but could have a small baseMean = mean of normalized counts. The independent filtering procedure takes care of this in a data-driven way, by guaranteeing an increase in sensitivity.

Another filter is to say, there should be some minimal amount of counts in some minimal number of samples:

filterGenes <- rowSums( counts(dds, normalized=TRUE) > x ) < y

This says, filter out gene where there are less than y samples with normalized count more than x.