Understanding log2 transformation for TPM

Question

[JACKY]

I have multiple RNA-seq datasets in a meta-analysis. All of those datasets are TPM normalized. However, one or two are also log2 transformed.

I don't understand what this transformation does, and how can it affect the analysis?

Can I still compare regular TPM to transformed TPM?

Quote from one of the articles:

" TPM values were log2 transformed for further analysis of individual genes or standardized gene pathway signature scores. Briefly, for each gene, we calculated the mean expression and s.d. across samples. Then, we subtracted the mean and divided by the s.d. to standardize the gene score to be centered at zero with units of s.d. (z-score)."

Thanks!

Answer 1

[Zhenyu Zhang]

in your quoted text, the first part is talking about log2 transformation, and the second part is talking about z-score transformation. They are not the same.

[JACKY]

Sorry I got confused. TPM log2 transformation and TPM, can I hold a comparison between them or are they not on the same scale?

What purpose does log2 transformation serve?

[Zhenyu Zhang]

Those 3 are different

• tpm
• tpm log2 transformation are log2(tpm) or normally log2(tpm+1)
• z-score