I am doing DE analysis using deseq2 and I noticed that the threshold of 'low count' changes when I have different inputs. Here' re my questions:
- How does deseq2 determine the low count threshold based on the raw count matrix? I can see they usually have NA in the normalized count matrix after deseq2.
- Do I need to apply an additional filter step when I select DE genes? Perhaps 'mean of normalized counts < 5' would be a reasonable criterion?
Honestly, I am a little confused. Peoples always use log2FC and adjusted p-value to select DE genes. Why low counts are never considered in this case? Do we just assume that the automatic filtering by deseq2 would be sufficient to filter out genes with low counts?