Should I average dN/dS over all CDS for a gene or consider only the longest transcript?

Question

[shreyasibiswas88]

Hi everyone,

I have calculated dN/dS ratio for ALL transcripts for ALL genes for my study systems (different species of rats) using codeml in PAML. Now I am not able to decide on what is the dN/dS for the entire gene (i.e all the CDS for a gene). Should I take an average of dN/dS ratio for all CDS of a gene or consider the dN/dS ratio of only the longest transcript of the gene as the dN/dS ratio for the that genes??

From what I see, dN/dS ratio between rat and mouse in ensembl is calculated as a cumulative ratio for all CDS of the gene.

What is the best practice when it comes to genome-wide dN/dS estimation? If someone can throw some light on this, it would really help me out. Thank you.