4.9 years ago by
United States
It's not really "noise", at least not according to the most common definition of the term. The issue is simply that isoform-level abundance estimation is often more difficult, largely because more multi-mapping occurs at the transcript level than at the gene level. If you imagine a gene with many similar alternatively-spliced transcripts, you can imagine that there will be a tremendous amount of multi-mapping between the isoforms of this gene, even if e.g. only one of these isoforms is expressed. Despite this added difficulty in assessing abundance as the transcript level, it is likely the case that, even if you wish to assess abundance at the gene level, you will likely get more accurate results by estimating transcript-level abundance, and then aggregating these abundances to the level of genes (see e.g. this article). Though it's not quite a measure of "noise", it's worth noting that a number of transcript-level expression tools provide the ability to perform bootstrap abundance estimates, or to draw Gibbs samples from the posterior distribution over abundance estimates, which does provide you with an estimate of the technical variability in the estimated abundances (see e.g. here or here).
•
link
written
4.9 years ago by
Rob ♦ 4.6k