Estimate alternative splicing rate using RNA-Seq

Question

[Nicolas Rosewick]

Hi,

I'm working on tumor sample (~20 samples), where three of them has a specific mutation in a gene involved in alternative splicing. We have RNA-Seq (stranded paired-end) for each of these tumors. What I want is to test if in these 3 mutated samples overall alternative splicing is dysregulated compared to non-mutated tumor. FYI we have also healthy samples.

My idea was to use MISO to make a simulation like this (MUT = sample with mutation, NO_MUT = sample without mutation)

1. MISO analysis : 3 MUT vs 17 NO_MUT
2. Count how many genes are differentially alternativelly spliced = MISO.real
3. Pick randomly 3 samples from the pool ( 3 MUT + 17 NO_MUT ) : known as MUT.sim
4. The rest of the 17 samples of the pool : known as NO_MUT.sim
5. MISO analysis : 3 MUT.sim vs 17 NO_MUT.sim
6. Count how many genes are differentially alternativelly spliced = MISO.sim[i]
7. Return to 3. and do it N=10000 times
8. p-value = number of MISO.sim >= MISO.real / N

What do you think ? Maybe is it a better way. And of course I don't know the computational time for one MISO run ...

Thanks

Answer 1

[Charles Warden]

I'm guessing you are doing this because MISO provides a Bayes factor rather than an FDR?

I would recommend trying MATS if you want a more traditional result table.