JunctionSeq: using continuous covariates, is it possible?

Question

[dhibar]

I'm wondering if anyone has any experience using continuous covariates in a JunctionSeq model for testing differential exon usage (DEU)? From the readJunctionSeqCounts function description it says: "design A data frame containing the condition variable and all desired covariates. All variables should be factors". Also, in the vignette all of the examples only involve factors.

My questions then are:

1. Is there some way to coerce the program to include numeric factors for things like age or components outputted from SVA?
2. If not, are there other programs that test for DEU that might allow for continuous covariates in the model?

Answer 1

[Charles Warden]

Unless you want to start altering the JunctionSeq code, I don't believe they've added the ability analyze continuous covariates.

For exon coverage, you could use DEXSeq. Or, if you use the QoRTs counts for JunctionSeq, you could analyze the exon and junction counts separately (so, junctions get a different dispersion value) with your counts-based differential expression algorithm of your choice, but you wouldn't have the visualizations provided by JunctionSeq.