Log transformation of RPKM data for PCA analysis

Question

[AB]

Hi everyone,

I have the RPKM values from my rna-seq dataset and I want to perform PCA on it. Taking log2 of the RPKM data gives infinity because of all the zeros which is not accepted by prcomp. Is there any other way to use the log transformation without removing the zeros?

Thanks in advance

Answer 1

[Tky]

You can just add a value such as 0.01 to all RPKM values for this situations. We routinely add 1 to the raw count value in our analysis.

[Ashutosh Pandey]

Adding 1 is more suitable as log2 of values less than 1 will give negative values.

[Tky]

Thanks for pointing out this, I thought it is the log ratio between the two conditions. Adding 1 make sense.