How can one perform regularized log transformation with RPKM values?

Question

[kajkajkajkajkaj]

Dear all,

I have RPKM normalized data and want to perform regularized log transformation with my matrix.

Would you please suggest me how to do it?

Best Regards,
Kaj

[Devon Ryan]

To my knowledge, this is only implemented in DESeq2, since it uses the shrinkage strategy employed by it. While you could try to simply input your matrix into the rlog() function, I suspect that this will result in an error (it will try to make a DESeqDataSet out of it and then likely complain about the fractional values). I suspect that your only recourse will be to either directly make the DESeqDataSet yourself so you can ignore those errors (and hope that nothing else in the rlog() function breaks) or to look at the code and reimplement the function. I would suggest the former route before resorting to the latter, due to the time involved.

[kajkajkajkajkaj]

Thank you very much, Devon.

However, I have no luck with transforming my matrix DESeqDataSet and try with rlog(). It ask me to input "Integers" (I think beacause it expects count). Alas, the second approach seems beyond the capability of biologist like me.

It would be very kind of you all to provide me some optional solutions.

[Devon Ryan]

Perhaps you can interest Michael Love in adding support for this. Since he wrote DEseq2, he'd be the quickest at making this happen. Aside from that, you might have to find someone willing to code things for you. Hopefully someone will reply with a pre-made package that can do this for you, but if not then those are pretty much your only options.

[kajkajkajkajkaj]

I will definitely try posting this in Bioconductor.

Thank you very much!