psi-coverage when 3 groups are compared

[M K]

Hello dear MAJIQ-Team,

I have a dilemma while constructing a pipeline for  MAJIQ steps, using 3 groups : WT, DKO, floxCTRL.

for #comparison-1: 

``` majiq-v3 psi-coverage      sg_WT_vs_DKO.zarr      WT_group.psicov     WT1.sj WT2.sj WT3.sj ```

for #comparison-2: 

```majiq-v3 psi-coverage     sg_WT_vs_floxCTRL.zarr     WT_group.psicov WT1.sj WT2.sj WT3.sj ```

Will these two commands produce a different <WT_group.psicov>, that I should prefix maybe with the comparison from which it 's coming ? Or I can skip this second comparison, since I have the WT_group.psicov and only run the command for the `floxCTRL.psicov` file ? 

Furthermore, the answer has a consequence in the Delta-psi step, where : 

``` 

 --splicegraph    sg_WT_vs_floxCTRL.zarr  \
  -psi1 WT.psicov \
    -psi2 floxCTRL.psicov} ...

```

In delta-psi, is any WT.psicov correct to use, produced either from the WT_vs_DKO.zarr or the WT_vs_floxCTRL.zarr ? 

This clarification is necessary when we build automated pipelines where all comparisons are run blindly.  

 I apologize if this question has already been answered already, but I only found examples with 2 groups in the previous questions and in the documentation site (https://biociphers.bitbucket.io/majiq-docs/v2-to-v3.html#v3).

I would appreciate any indication or documentation on this, 

Have a nice day,

with Regards,

Maria

[San Jewell]

Hi Maria, 

I'm slightly confused by the pipeline, but I think hopefully it will be an easy clarification. Basically, if downstream you're going to anticipate running a comparison between multiple psi-coverages using deltapsi or heterogen, these will need to be compared over a unified splicegraph. It is not possible to run a delta-psi between two psi-coverages which may have different splicing patterns in their respective splicegraphs. LSVs need to be built over all structures so that they are harmonized. So, instead of having sg_WT_vs_DKO.zarr and sg_WT_vs_floxCTRL.zarr, there should just be one splicegraph.zarr, with a build comprised of all groups. I think then the rest of your questions are answered. (note for efficiency, it is technically possible to add new samples to groups without needing to recalculate all lsvs, which is what majiq-clin takes advantage of, but we are still deciding on an appropriate paradigm to propose to users before making a recommendation) 

Let me know if it makes sense, 

-San