Using EBSeq with normalized counts and PCA

[Pankaj Agarwal]

Hi,

As per the EBSeq vignette non-normalized counts are needed to use EBSeq for gene expression analysis.  But I have TPM counts that are normalized.  Can I still used EBSeq and what would I have to do to take into account that the counts are normalized.

Further, does EBSeq provide any function for PCA?

Thanks,

- Pankaj

[rik.ve...@gmail.com]

Hi Pankaj,



EBSeq fits a negative binomial distribution on the raw counts. After normalization, your data will not be NB distributed anymore, so the models won't fit as they should. Do you have info about the total library sizes? In that case you can back-calculate the counts, and use EBSeq procedures as outlined in the vignette.

I have not found PCA functions in EBSeq, but I don't see why the standard prcomp function in R would not do...

Best,
Rik