MAJIQ data interpretation problem

[Harisankar Harikumar Sheela]

Hi Biociphers team ,

My name is Hari, and I am a graduate student in the Barmada lab at the University of Michigan. I am writing to you as I had some trouble running MAJIQ and Voila. I had taken RNA seq datasets and analyzed splicing changes using the above-mentioned software. I was able to identify genes that underwent splicing changes and the regions of splice variation accurately. However, the problem I ran into was that the changes that the software predicted were in the exact opposite direction from what I observed in the biological data. I double-checked my entire methodology carefully and could not find any errors in the way I had performed the analysis. I was wondering if you may have any insights on this and why this may be the case. 

I greatly appreciate the help with this matter. Thank you so much for your time and support.

Sincerely, 

Hari

--

Harisankar Harikumar Sheela

PhD candidate | Department of Molecular, Cellular, and Developmental Biology (MCDB)

University of Michigan

[Matthew Gazzara]

Hi Hari,

Would you mind sharing some example data and what your expectation is for the changes based on your other biological data? Something like screenshot(s) of the Voila output for an event of interest or some of the TSV output and what your expectation is for the RNA-seq results? With that I can help interpret the results.

It could be something like switching "group A" and "group B" when you run MAJIQ that could give you dPSI values of the opposite direction than what you expected. Also since we give splice junction usage values for all junctions involved in an LSV sometimes looking at the "wrong" junction could give you opposite values compared to expectations. For example, for a cassette exon you may see in the literature the inclusion level to increase from condition A to condition B, but if you focus on the splice junction for exon skipping this could have a negative dPSI value from MAJIQ.

-Matt