Hi all,

I would like to compare, statistically, the expression levels of orthologous genes across different species. The main issue that arise with this is that TPM may be influenced by gene length and, because orthologs are not necessarily of the same length between two species, I think I cannot directly do stats to compare their TPM levels without being biased. Is the use of LengthScaledTPM done exactly for this particular situation? Meaning that if I get 'CountsfromAbundance' (tximport package) in terms of LengthScaledTPM, have I discarded the influence of gene length on the measure of expression level, so that I can compare them across species? Said otherwise, can I run stats on LengthScaledTPM values to compare them across species without concerns for biased estimate due to different gene length?

Many thanks in advance for your answers, Guillaume

To start, TPM is designed to not have length bias. It divides out the length from the observed count.

However, any of the code chunks about handing off data to statistical packages, outlined in the tximport vignette, will correct for sample-specific length differences. So as long as you follow any of those workflows, you will be correcting for gene length across samples.