What is the Difference Between Within Gene Isoform Frequency and Global Isoform Frequency?

Question

[jack]

Hi all,

I really didn't understand the difference between gene isoform frequency and global isoform frequency. Can anybody explain what is it exactly?

[Devon Ryan]

Can you give an example of where "global isoform frequency" is used? I'm not sure how to interpret that out of context.

[jack]

For each simulation set, we computed abundance estimates with the tested methods and measured the accuracy of the transcript fraction estimates using the median percent error (MPE), error fraction (EF), and false positive (FP) statistics that we used previously [7]. The MPE is the median of the percent errors of the estimated values from the true values. The 10% EF is the fraction of transcripts for which the percent error of the abundance estimate is greater than 10%. Lastly, the FP statistic is the fraction of transcripts with true abundance less than 1 TPM that are predicted to have abundance of at least 1 TPM. These statistics were calculated for three levels of estimates: (1) gene relative abundances, (2) global isoform relative abundances, and (3) within-gene isoform relative abundances. http://www.biomedcentral.com/1471-2105/12/323

[Devon Ryan]

Hmm, it's still a bit unclear even after reading the RSEM paper. They only use "within isoform frequency" in the context of talking about Miso, which I've never used. RSEM/Cufflinks/etc. output RPKM/FPKM metrics for each isoform or gene, which would seem to be a reasonable example of global isoform frequency. My only guess would be that Miso doesn't (or at least previously didn't) do that, perhaps instead outputting a percentage of each isoform (e.g., within gene "Foo", isoform A is 30%, isoform B is 50% and isoform C is 20%). Hopefully someone who's used Miso will end up reading this and be able to provide more insight.