Hi,

I am using DeSeq2 for a set of metatranscriptome data for which I also do have estimations for the number of total transcripts per L within each sample. As I can clearly see from my data that the number of transcripts per L is influenced by the type of treatment, I would prefer to make statistics based on absolute instead of relative expression data. My idea was to multiply the automatically generated DeSeq2 size factors with a vector reflecting the number of transcripts per L in each sample (set to 1 for the sample with the lowest number of transcripts per L). Does this make sense and is it statistically valid to work with such manipulated size factor?

Best regards,

Sara

This approach doesn't make sense to me. If I knew there were a big discrepancy in the number of genes expressed across condition, to the point that the size factors might not be accurately estimated, my first try would be to use the controlGenes argument of estimateSizeFactors(), to limit the size factor calculation to the intersection of expressed genes across condition.