I'm using org.Hs.eg.db to annotatate ENSEMBL gene IDs from a read counts matrix, and after my list is created, I have ~3700 that have 'NA' for the annotation. If I keep these unannotated genes, perform DE testing in edgeR, and retrieve my topTags, I see quite a few of the unannotated genes in this list. Some of these genes are lincRNA (ENSG00000205959, ENSG00000260743, etc.) pseudogenes (ENSG00000236297, ENSG00000232901, etc), and antisense RNA (ENSG00000267734).
How do I deal with these genes? Should I filter to remove any genes that do not have annotations prior to normalization and DE testing or should I leave these in the analysis? Obviously, it makes adding annotations and generating figures cleaner, but I'm not sure what the convention is. Thanks.
They're just that, unannotated. I'd keep them, may represent novel findings.
Thanks. I figured as much, but I've seen research removing data that doesn't contain official gene symbols: https://f1000research.com/articles/5-1438/v1. Just curious what the consensus is on this type of filtering.