I have a data set where I am comparing a few exposures to a basleine group. When I run Deseq on all of the groups it filters out about 20,000 of the 36,000 genes which is what I expect and has always been the case when running Deseq. However, for 1 of the contrasts only 200 genes get filtered out. Looking at the results table it looks like almost 15,000 genes that make it through have counts < 1. Does anyone know why the low counts filter would not be filtering here? 

out of 35599 with nonzero total read count

adjusted p-value < 0.1

LFC > 0 (up)     : 0, 0%

LFC < 0 (down)   : 0, 0%

outliers [1]     : 0, 0%

low counts [2]   : 205, 0.58%

(mean count < 0)

[1] see 'cooksCutoff' argument of ?results

[2] see 'independentFiltering' argument of ?results

[1] "Genes < 0.05:"

[1] 0

The filtering is an optimization based on how much power is gained by filtering at a certain threshold. If users have many pairs to compare it may be more consistent to pick a common threshold and set independentFiltering=FALSE. You can filter at the beginning based on mean of normalized counts.