voila issues and doubts

Miriam Martínez

unread,

Oct 22, 2025, 1:49:11 AMOct 22

to Biociphers

Hi all,

I'm currently having some issues when running voila (both tsv and view) and was wondering if I could have some help.

Regarding voila view, the other way I was able to run it and see my data but the different options for showing the splicegraph and changing the options displayed, were not accessible (they were in the html, but it was like if the links were broken as it didn't appear any options, except for the index, that made me return to the lsvs list). Now, I'm trying to run voila view with the same files, but it gets stucked on the voila_index.hdf5 step (already tried to force the index and gets stuck on the same step).

Regarding voila tsv files, I'm still running some tests as I'm having issues getting any results, so I will come back to this once I have some results (or left enough time to check if there is any result in the output). However, there is one thing that I would like to ask if it is normal or this shows that something is not correct: on all the tsvs that I'm obtaining from voila tsv, on the header all indicate "psi_samples:no metadata" and also in "test_percentile". Is this normal or there is something that the tool is not selecting correctly or that I'm not specifying? Also, I have two questions about the options on voila tsv:

1) If you have more than one statistic on your results, which statistic does voila tsv check for significance? Or does the LSV have to be significant for every statistic selected in order to pass the pvalue filter?

2) When using the --show-all option, does the output reflect all the LSVs that are over the --changing-between-group-dpsi N value, or it actually shows ALL the LSVs?

These are my questions right now and once I have a more clear picture of the results I'm obtaining in my different test runs, I'll update my questions regarding the tsvs. Thank you very much in advance!

Best,

Miriam

San Jewell

unread,

Oct 22, 2025, 12:04:45 PMOct 22

to Biociphers

Hi Miriam,

I will try to separate this into the questions I can determine:

q1) Voila view has clickable buttons but they do not do anything when clicked: As this is not an issue I have observed yet, can you please let me know how to reproduce the issue, which version of majiq/voila you are using and what are the steps to make the issue occur? (From the workshop example data if possible)

q2) Voila indexing gets stuck: There may be a few causes for this. Again can you confirm which voila version you are using? Which kind of analysis (psi, multiple psi, dpsi, het) you are supplying to voila / the list of switches for voila? Sometimes the message "listening on port XXXX" will not display, but the voila web server is still running. Have you tried connecting to it as you normally do? When the process gets stuck, is there anything obviously working/running? For example, can you see any processes on your machine for voila using a lot of disk or a lot of cpu usage in your system monitor?

q3) The missing metadata is at this time expected in v3. There weren't similar attributes as in v2 and instead of removing the v2 keys they have just been listed as "no metadata". It's likely they should just be removed entirely but the decision was not finalized in the lab for now. They can be safely ignored and do not indicate that there is anything wrong with how you have run the analysis.

q4) (your question (1)) , the stats are used in het mode for determining the values of the changing and non_changing columns. As the --help text states for these:

--non-changing-pvalue-threshold Uses minimum p-value from all tests provided.

--changing-pvalue-threshold Uses maximum p-value from all tests provided.

q5) (your question (2)) ; for heterogen mode, the lsv list is not filtered based on threshold/probability-threshold ; these are for dpsi mode only, and as such, --show-all only makes a difference for dpsi mode. This should be better clarified in the --help text. At the moment for het it is expected to filter and sort by changing/non-changing after the tsv is generated but more options can be added as well to tsv mode. If you have recommendations, we will consider them.

Please let me know if my responses make sense.

Thanks,

-San

Miriam Martínez

unread,

Oct 22, 2025, 10:50:18 PMOct 22

to Biociphers

Hi San,

I will answer with the separation you used on your last response:

q1) My voila version is: voila 3.0.10.dev1+g6cdae2a0f (still waiting the managers of the cluster I work with updates it with the --show-reads and --show-psi fixes). Just to specify, the clickable buttons that don’t work for me are the “Splice Graph”, “Plot Options” and “Toggle Splice Graph”. I will try to reproduce the issue with the workshop data, but it will take me some time to do so. Another option is that I send you the files that I’m using to run voila (as well as the command I’m using or that I have used) to check if it also happens when you run it. Please let me know if you prefer this option or if it is viable (I don’t know if just sending you the sg.zarr and the .sgc files is enough), and if so, which commands should I send you (from which step and on).

q2) I’m doing heterogen analysis, and as I’m working on a cluster, I don’t have much control of the resource allocation nor monitoring. Nonetheless, I’ve tried opening the localhost and it worked, it just didn’t appear the "listening on port XXXX". Even though it did open, I’m not sure if the filters are displaying correctly. I’ve attached 2 images to explain this case: on voilaview1, I have set the thresholds of abs(E(dPSI)) on 0.1 and p-value on 0.05 but, checking voilaview2, it can be seen that the event shown on the first picture should have not passed the abs(E(dPSI)) filter (delta psi is -0.038). So, I don’t know what exactly is happening here (also, the stat is infoscore on the picture because I can’t change the display options as I mentioned before, but I will try to reproduce the error).

q3) Great, good to know!

q4) Ah sorry, seems I must have skipped the --help text for these options when reading. Thanks for the clarification.

q5) Ok! I thought the --show-all applied to all modes. So, if I have understood correctly, when using heterogen analysis there is no dpsi filter applied right? On the tsv all the analyzed lsvs are shown and then that tsv should be filtered with the help of the changing and non-changing columns.

Regarding recommendations on this specific point, I don’t know how viable/difficult would be to include a dpsi column that indicates the dpsi for each event on the lsv (just as it is shown when using voila view). In my experience, the filtering would be more straightforward as it would probably need less formatting to check which events of interest are over a particular dpsi threshold, specially when you are doing an exploratory or genome-wide analysis and not looking for a specific gene or group of genes. Hope this makes sense.

Thanks for your reply and I will let you know if I’m able to reproduce the issue with the workshop data.

Best,

Miriam

voilaview1.png

voilaview2.png

Miriam Martínez

unread,

Oct 23, 2025, 1:35:40 AMOct 23

to Biociphers

Update to last message:

I have done the analysis with the same scripts and options with the workshop data and the clickable buttons work perfectly fine, so I can't reproduce the problem with this dataset. Maybe the issue is due to my high N on the analysis?

If I can do another thing to try to find where the problem is or you need any info please let me know (if you need any files related with my analysis/samples, I'd rather send them via a private message). Thanks again!

Best,

Miriam

San Jewell

unread,

Oct 23, 2025, 2:59:35 PMOct 23

to Biociphers

Thank you for the detailed response!

1) + update , I would guess there is something in the data or way it's ran, but it's likely it's not actually the contents of the bam files, etc, probably just something not accounted for in the way the analysis was run. Instead of the data files, would you be able to provide the bash script, and if applicable, the config file used for the majiq steps plus voila step? IT might be that if I arrange a similar sequence, but with workshop data instead, I might be able to reproduce the error. If I cannot, then I would probably need to work with you to get the data itself.

2) I was able to reproduce this one. It seemed that the data for the max dpsi was never even passed to the index, so no filtering was happening on the index page, oops! I pushed another minor version bump for this. It would require running voila with --force-index to regenerate the index.

5) You are correct that there should be no quantified lsv rows missing from the tsv output for heterogen run mode. Let me know if anything seems amiss here or if you are seeing anything different.

Thanks,

-San

Miriam Martínez

unread,

Oct 24, 2025, 12:05:30 AMOct 24

to Biociphers

Hi San,

1) I have created a file that groups all the commands I run (I usually run each step separately), including the voila view command. Also, I will send you the config_file and the model_matrix file I used for moccasin too in case it is useful (I will send you all these files on a separate mail as I'm not sure if it is fine that I share the config_file/model_matrix in public and I can't ask my supervisor right now). Please, if you don't get the mail with the files, let me know and I will resend them.

2) Glad it was solved!

5) Ok, thanks for confirming! For now I'm still running some tests with the voila tsv command. As I have a really high N, it takes nearly nearly 18-19h to generate a tsv file without including reads or psi per sample (I don't know if this time could be reduced by assigning more memory to the process in the cluster).

Thank you!

Best,

Miriam

Maximo Rivarola

unread,

Oct 25, 2025, 3:37:17 PMOct 25

to Biociphers

hi everyone,

thanks for a great tool!! I am almost there, .... I have an Issue

Thanks in advanced for any help!! :)

voila tsv file inputs???

Voila Execution Problem Summary

We are unable to generate reports using voila tsv or voila view with the outputs from MAJIQ v3. (running on server) installed with micromamba:

Software Version: MAJIQ and Voila version 3.0.13.dev1+g2db64969d. Input Files: Valid Zarr format files generated by MAJIQ v3:

splicegraph.zarr (from majiq build) .voila files (from majiq deltapsi) all_samples.psicov.zarr (from majiq psi-coverage

Persistent Core Error

voila -h and voila tsv -h help messages, and providing valid input files (including testing absolute paths), both voila tsv and voila view consistently fail with the error:

"ERROR - Could not find any .sgc experiment data inputs Traceback (most recent call last): ... Exception: Could not find any .sgc experiment data inputs"

From what I and chatGPT can understnd:

The error message mentioning .sgc files appears misleading, likely referencing older MAJIQ v2 formats. The fundamental problem is that this specific development version of Voila (3.0.13.dev1+g2db64969d) seems unable to correctly recognize or parse the splicegraph.zarr directory??? is this correct?

my dir and files:

[mrivarola@serafin 06_majik311]$ ls 01_majiq_build_output/splicegraph.zarr/
contigs exons genes introns junctions zarr.json

[mrivarola@serafin 06_majik311]$ ls 01_majiq_build_output/T1_Control_1_Aligned.sortedByCoord.out.sj/
contigs sj_introns sj_introns_bins sj_junctions sj_junctions_bins sj_junctions_contigs zarr.json

command:

micromamba run -n majiq_py311_np1 voila --license /home/shared/epilab/TomateOmics_all_Nov2024/Priming_RNA_seq_2025france/06_majik311/licenseMajiq.lic tsv \
/home/shared/epilab/TomateOmics_all_Nov2024/Priming_RNA_seq_2025france/06_majik311/01_majiq_build_output/splicegraph.zarr \
/home/shared/epilab/TomateOmics_all_Nov2024/Priming_RNA_seq_2025france/06_majik311/04_majiq_deltapsi_output_v3/T2_C_vs_T2_P.voila \
/home/shared/epilab/TomateOmics_all_Nov2024/Priming_RNA_seq_2025france/06_majik311/02_all_samples.psicov.zarr \
-f /home/shared/epilab/TomateOmics_all_Nov2024/Priming_RNA_seq_2025france/06_majik311/T2_C_vs_T2_P_abs_path_psicov_test.voila.html \
--threshold 0.1 \
--changing-between-group-dpsi 0.1 \
-j 6

2025-10-25 15:28:08,411 (PID:2293420) - INFO - Command: /home/mrivarola/micromamba/envs/majiq_py311_np1/bin/voila --license /home/shared/epilab/TomateOmics_all_Nov2024/Priming_RNA_seq_2025france/06_majik311/licenseMajiq.lic view /home/shared/epilab/TomateOmics_all_Nov2024/Priming_RNA_seq_2025france/06_majik311/01_majiq_build_output/splicegraph.zarr /home/shared/epilab/TomateOmics_all_Nov2024/Priming_RNA_seq_2025france/06_majik311/04_majiq_deltapsi_output_v3/T2_C_vs_T2_P.voila /home/shared/epilab/TomateOmics_all_Nov2024/Priming_RNA_seq_2025france/06_majik311/02_all_samples.psicov.zarr -j 6

2025-10-25 15:28:08,411 (PID:2293420) - INFO - Voila v3.0.13.dev1+g2db64969d
2025-10-25 15:28:08,413 (PID:2293420) - INFO - ╔═══════════════════════════════════════════════════════════════╗
2025-10-25 15:28:08,413 (PID:2293420) - INFO - ╠╡ ACADEMIC License applied ║
2025-10-25 15:28:08,413 (PID:2293420) - INFO - ║ Name: Official Majiq Academic-only License ║
2025-10-25 15:28:08,413 (PID:2293420) - INFO - ║ File: licenseMajiq.lic ║
2025-10-25 15:28:08,413 (PID:2293420) - INFO - ║ Expiration Date: Never ║
2025-10-25 15:28:08,413 (PID:2293420) - INFO - ║ ║
2025-10-25 15:28:08,413 (PID:2293420) - INFO - ╠╡ The academic license is for non-commercial purposes by ║
2025-10-25 15:28:08,413 (PID:2293420) - INFO - ╠╡ individuals at an academic or not for profit institution. ║
2025-10-25 15:28:08,413 (PID:2293420) - INFO - ╚═══════════════════════════════════════════════════════════════╝
2025-10-25 15:28:08,413 (PID:2293420) - INFO - config file: /home/mrivarola/tmp/voila-config-da1ce2afb70c98be7a2de6ec3c79a6e7
2025-10-25 15:28:08,649 (PID:2293420) - ERROR - Could not find any .sgc experiment data inputs
Traceback (most recent call last):
File "/home/mrivarola/micromamba/envs/majiq_py311_np1/lib/python3.11/site-packages/rna_voila/run_voila.py", line 526, in main
config.write(args)
File /home/mrivarola/micromamba/envs/majiq_py311_np1/lib/python3.11/site-packages/rna_voila/config.py", line 315, in write
sg_file = find_splice_graph_file(args.files, getattr(args, 'splice_graph_only', False))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/mrivarola/micromamba/envs/majiq_py311_np1/lib/python3.11/site-packages/rna_voila/config.py", line 147, in find_splice_graph_file
raise Exception("Could not find any .sgc experiment data inputs")
Exception: Could not find any .sgc experiment data inputs

thanks again!!!

cheers, maximo,

any ideas or help are greatly apreciarted

Miriam Martínez

unread,

Oct 26, 2025, 8:30:40 PMOct 26

to Biociphers

Hi Maximo,

I think I can help you with these issues. In order to execute the voila command (tsv and view) you need the following:

1) splicegraph.zarr (from build)

2) the .voila file (from psi/deltapsi/heterogen analysis)

3) the .sgc files (from sg-coverage)

Reading your output, seems you are missing the .sgc files, which are a separate step that has to be done before running voila tsv. For creating a .sgc file you need to run:

majiq-v3 sg-coverage path_to_build/splicegraph.zarr path_to_output_sgc_file/name_sgc_file.sgc (this is just for naming your .sgc file) file1.sj file2.sj (and so on with all the .sj that are part of the group).

Nevertheless, they have a really detailed documentation and on one of the sections they have a nice migration guide between majiq v2 and v3 that has an example with the different steps for majiq for both versions in case it is of help too (https://biociphers.bitbucket.io/majiq-docs/v2-to-v3.html).

Hope it was useful!

Best,

Miriam

Maximo Rivarola

unread,

Oct 26, 2025, 10:14:58 PMOct 26

to Biociphers

HI Miriam, thanks for clarifying, very clear now!

I am running as you suggested, seems to be working great, specially with majiq-v3 command, not just majiq.

thanks again!!! will keep reading docs and will let you know how it goes, :)

thanks again!

cheers, maximo

San Jewell

unread,

Oct 27, 2025, 4:51:07 PMOct 27

to Biociphers

Hi Maximo,

I am glad that Miriam was able to answer your question.

I would ask that you kindly make a new thread in this group when you are asking a new question. Please do this for any future posts you make here.