What to do with genes with very low counts in a dataset

[tellsparck]

Hi all,
I have a dataset from a single cell RNASeq project and there are many genes with very few counts (1 or 2 in a few samples) in the data set. While this is important info (tells us that these genes are not expressed), how would one handle them for a differential expression analysis (I use DESeq2)? Should I delete them or keep them? For me this means the difference between having a dataset with either ~17000 genes or ~11,500 genes (if I exclude genes with an average count less than 3).
Thanks for your thoughts on this!

[dpryan]

DESeq2 will filter them out for you, you needn't do it manually.

[Michael Love]

Devon's always got the right answer

If you want to know more about filtering out genes with few counts, we have a section on this (independent filtering) in the "Theory" part of the DESeq2 vignette.

A reference on independent filtering:

Bourgon R, Gentleman R, Huber W.
"Independent filtering increases detection power for high-throughput experiments."
Proc Natl Acad Sci U S A. 2010 May 25;107(21):9546-51. doi: 10.1073/pnas.0914005107. Epub 2010 May 11.

http://www.ncbi.nlm.nih.gov/pubmed/20460310

[tellsparck]

Thanks for the quick replies, dpryan and Michael Love!