rMATS-long for 3' vs 5' capture libraries

[ethieme]

I know that most long read technologies rely on polyDT 3' capture, but I am currently using 5' capture and I'm wondering whether there is anything inherently built into the rMATS-long model that I should be aware of before using this to look for differential isoform usage in my 5' data? I plan to just use my own standard pipeline for sort/align/collapse isoforms to get a known + novel trx GTF, so I would only be using rMATS-long for the actual statistical model to check for differential isoform usage.

[EricKutschera]

rMATS-long requires a GTF file that defines the transcripts and an abundance file which gives the read counts for each transcript in each sample. Those files can be from ESPRESSO like in the example data:
https://github.com/Xinglab/rMATS-long/blob/v1.0.0/example/samples_N2_R0_updated.gtf
https://github.com/Xinglab/rMATS-long/blob/v1.0.0/example/samples_N2_R0_abundance.esp

You can use your own pipeline to get transcripts and counts for input to rMATS-long

When rMATS-long processes those input files it's not assuming anything about 5' or 3' bias. The statistical test doesn't use the transcript structure. After the statistical test, the GTF is used to make plots and to report alternative splicing differences between pairs of transcripts

[ethieme]

Hi Eric,
Thank you for the prompt and thorough reply! Thats what I thought, thanks for confirming before I dive in. I appreciate your work building/maintaining this tool and I'm excited to see what comes out!
Best,
Elana