Statistics for within-condition transcript abundance comparisons
1
0
Entering edit mode
nicog • 0
@nicog-16857
Last seen 2.2 years ago

Many of the tools I've used provide insight into differences in splicing events or DTE/DTU between conditions, but what if I want to make comparisons (with some measure of confidence) between isoform abundances within conditions. Could I just use the estimated abundance of each transcript from Salmon quantification corresponding to a gene of interest? If so, what statistical test would be appropriate for the comparison? The null hypothesis, here, is that all expressed (vs. annotated?) isoforms contribute equally to the total gene abundance. Would an Anova + (paired?) Student's t-test with some kind of p-value correction be valid here?

Thanks, Nico

rnaseqDTU • 265 views
2
Entering edit mode
@mikelove
Last seen 12 hours ago
United States

My preferred approach would be to estimate the proportion vector p within a condition with either posterior SD or SE on the components of p.

Given that the null hypothesis is trivially false for many (all?) genes, I think the p with bands on each component is sufficient for describing one gene.

I think there should be a way to propagate from e.g. DRIMSeq estimated precision to the uncertainty on p. I'll try to bring this thread to the package authors' attention.

0
Entering edit mode

Simone from Mark Robinson's group has an answer for you:

BANDITS::plotproportions outputs estimated mean transcript relative abundance (pi1, ..., pi_K) for each group of samples, with Wald-type CIs.