How are you defining gene/transcript expression?

[blahah]

In rna-seq.tex you have excellent discussion of various aspects of expression quantification. I think there are some missing pieces:

• Given that you state you are interested in transcript quantification, how do you actually quantify transcripts? Mapping to genomic features does not provide this because of shared exons, so how are you getting transcript abundance measures? And how do you handle multi-mapping reads?
• In several places you mention gene expression. How are you defining this? If the molecule of interest is the transcript, how is gene expression defined relative to expression of each transcript?
• RPKM is unstable across samples by definition because the normalisation factor is sample-specific. This is not the case for TPM, which is only unstable due to the molecular population makeup instability that you mention.

[klmr]

Ooh, excellent points. I should definitely be careful to use a consistent terminology. And as for the gene–transcript quantification, I used some rather ugly simplification back when I did this, which consisted of quantifying gene expression and assigning it to the “canonical” transcript (the longest annotated CDS). Terrible, and I wouldn’t do it like that today, but in terms of codon usage it actually performs the task with very little bias.

Thanks, I’ll amend/explain that.

[blahah]

I was gonna read the rest of the thesis when I have time - let me know if this would be helpful now or if it's too soon to make useful comments.

[klmr]

I was actually going to submit immediately after BoG, so it’s definitely not too soon. The thesis isn’t extensively proofread yet but apart from that (and the lack of supplements) nothing much will change. If it helps you I can give you the reference library so you can compile the document — or you may prefer having the compiled document?

[blahah]

Great! Happy with either solution :)