How to calculate pvalue, FDR and IncLevelDifference

[lsh74]

Hello, I feel very confused after reading the source code. If I have two groups of samples, and the PSI value of each sample forms a count matrix, how can I calculate the three values of pvalue, FDR and IncLevelDifference? It would be great if you could help me solve the problem

[EricKutschera]

The PSI values (IncLevel1, IncLevel2) are calculated based on the read counts (IJC_SAMPLE_1, SJC_SAMPLE_1, IJC_SAMPLE_2, SJC_SAMPLE_2) and the effective lengths for the event (IncFormLen, SkipFormLen). Here's the code to calculate a PSI value: https://github.com/Xinglab/rmats-turbo/blob/v4.2.0/rMATS_P/inclusion_level.py#L24

That code is essentially: (IJC_SAMPLE_1/IncFormLen) / ((IJC_SAMPLE_1/IncFormLen) + (SJC_SAMPLE_1/SkipFormLen))

IncLevelDifference is average(IncLevel1) - average(IncLevel2). Here's the code: https://github.com/Xinglab/rmats-turbo/blob/v4.2.0/rMATS_P/inclusion_level.py#L65

The code for calculating the PValue is here https://github.com/Xinglab/rmats-turbo/blob/v4.2.0/rMATS_C/src/myfunc.c#L462

The details are described in the supporting information file of http://dx.doi.org/10.1073/pnas.1419161111, but the code implements a likelihood-ratio test for the difference in PSI value between the two groups exceeding some threshold

Here's the code for calculating the FDR from the list of PValues: https://github.com/Xinglab/rmats-turbo/blob/v4.2.0/rMATS_P/FDR.py