Entering edit mode
3.4 years ago
Rahul Tyagi
▴
10
Hi all,
I read that MAST can be used with batch information to properly find DE genes, however I can't quite find how this can be done. I have used MAST with Seurat before, but with no batch information. Can someone tell me how this can be done?
Hello, MAST seems to take a design formula (like any other model-based framework such as DESeq2, edgeR, etc). Here is an example. This is the code run for the benchmarking paper from Soneson and Robinson (2018).
You would add the batch information as a so-called covariate to the design. This https://rdrr.io/bioc/MAST/man/zlm.html is the function that takes the design, and this you then feed e.g. into https://rdrr.io/bioc/MAST/man/lrTest.html
I am not a MAST user, maybe they have alternative tests beyond the LRT, you would need to consult the manual http://rglab.github.io/MAST/articles/MAST-Intro.html
Does that all make sense to you?