Hello,
I am happy to see that Limma is updated to handle RNA-Seq data. I have FPKM values for my transcripts. Can I use them directly as input to Limma? Is there any kind of normalization to do on FPKM values prior to using Limma?
Thanks,
Delasa
Limma has been upgraded to use RNA-Seq COUNT data, not FPKM. You could use 'regular' limma analysis if you take logs and use trend = TRUE in your call to eBayes(), but otherwise most of the available analysis packages in Bioconductor expect counts. See A: Differential expression of RNA-seq data using limma and voom() for more information.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
Thanks for your response. Just a quick question:
Is taking logs a necessary step for this analysis (if I choose to use Limma for FPKM values)? That would result INF in case of FPKM = 0. Is that internally handled in Limma?
Yes, you would use limma on log FPKM values. One typically adds a small constant to all the values before taking the logarithm to avoid taking the log of zero. Again though, FPKM is not the ideal input data for a differential expression analysis. See the link in James MacDonald's answer.
You can just add a small constant to your FPKM values (say 0.25) to account for the zeros.