-
Notifications
You must be signed in to change notification settings - Fork 1
Description
junctionCounts version 1.0.0
In one of the responses to issue 17 it was noted that:
Significance of events is best established by providing all events and event types. Due to the nature of DEXSeq's dispersion estimate and comparison procedures, the less events you input, the fewer it calls as significant.
By the same token, it seems that including more samples would also improve calling.
Now in my work with DESeq, I supply read counts for all samples to generate dispersion estimates and sizeFactor calls. Then I use the program to call significant events for various pairwise comparisons of their conditions.
A typical wetlab experiment consists of more than 2 conditions. Typically, there may be one control condition and multiple treatment conditions. Or there may be a time series of conditions. In either case there are multiple comparisons of interest.
I would request that the same be implemented for junctionCounts DEXSeq. A single sample table could include all the samples, assigning them to various conditions (as in the present implementation). A new parameter would indicate the condition-pairwise comparisons desired.
Thanks for your consideration.
Sol Katzman
UC Santa Cruz Genomics Institute.