zarr.errors.FSPathExistNotDir

[M K]

Hello dear MAJIQ Team,

I have an issue with my script, at the "deltapsi" step, but it may originate from "build".

I attach you here the log file with the error, after running in a Slurm-cluster.

Also you can see my script, with all the commands.

Could you answer me to the following questions, aside of helping me understand the error message I received. 

- Do I need to do run the "sj" step for all the samples ? 

- Should I get a .zarr folder output after the "build" step ? 

- Is it correct to run a delta-psi for 1 sample-vs- all the others, when my experiment consists of patient data where each sample may have one causant splicing event ?

- Which command should I adapt to answer to this kind of comparison and obtain the "deltapsi " ? > psi-coverage, build, or other ?

Thanks a lot for your time and any valuable insight,

Best regards, 

Maria

Institut of Myology, Paris

[M K]

...I also add the config-file as used in my script above. Sorry for omitting!

[San Jewell]

Hi Maria, 

The issue here with the "majiq-v3 deltapsi" command is that the flags "-psi1" and "-psi2" are meant to point towards .psicov dirs (result of majiq-v3 psi-coverage) , but in your command they are pointing to .tsv files I believe if you changed in your script file lines 194 and 195 like so, it should work:

  -psi1 ${work_dir}/psi_cov/K_vs_all.psicov \

  -psi2 ${work_dir}/psi_cov/all_vs_K.psicov

Let me know if it helps!

-San

[M K]

Hello San,

thanks a lot for your response! That was eventually the error ! 

But I am still confused about the fact that the error mentions the .zarr file, because I didn't really obtain it in the build step. 

Is this normal ? Has the output changed in  v3 ? 

According to tutorial in bitbucket.io page : 

Output files

• SpliceGraph file (splicegraph.zarr) defining annotated and novel exons...

But, my output after the build step is the following: 

[../build_output$ ls  1sample_vs_all/C002I3K_vs_all/]
[contigs  exons  genes  introns  junctions ]  

...where each folder contains only '0',' .zarray', '.zattrs'.

--> Finally, for visualizing the results, could I use voila, after the deltapsi ?

Following this example in the bitbucket.io tutorial again : 

1.[ majiq-v3 sg-coverage /path/to/results/build/sg.zarr /path/to/results/build/Brain_Cerebellum.sgc /path/to/results/sj/sample_bc_1.sj /path/to/results/sj/sample_bc_2.sj ] 

2.[ voila view /path/to/results/build/sg.zarr /path/to/results/psi/Brain_Cerebellum.psicov /path/to/results/build/Brain_Cerebellum.sgc ]

..I don't know what to put as  "/path/to/results/build/sg.zarr". The whole folder of my build-output would be the same ?

Could you clarify me this please? 

Thanks a lot for your help,

Have a nice day,

Best regards,

Maria

[San Jewell]

Hi Maria, 

To clarify, in v2 of majiq, to store the splicegraph data we used splicegraph.sql which was sqlite format, in v3 we use zarr format. Zarr is a general purpose library which supports a number of different ways of storing the data on disk. Some of them create a single file for all of the data, but the one currently used in majiq v3 leverages the file system for creation of the tree structure of internal data, which means it is not actually a file but rather a folder/directory. (inside the folder/directory are other directories "contigs  exons  genes  introns  junctions" as you mention) By convention when doing a build the splicegraph directory in the output will be called "splicegraph.zarr" which I would encourage you to stick to using for clarity, however, technically you can name this directory anything you want. 

In this case voila is looking for the path to 1sample_vs_all/C002I3K_vs_all for splicegraph. 

Let me know if it makes sense. 

Thanks, 

-San