HDF5 dataset ".o5_o3_psi5_psi3_dimnames/2" already exists

[ChiaraF32]

Hi,

I have been having this issue when I rerun the aberrant splicing module (e.g. to add more samples to the run):

Error in .normarg_h5dimnames(h5dimnames, group, not_NULL, filepath, name): HDF5 dataset ".o5_o3_psi5_psi3_dimnames/2" already exists

This occurs during the rule: AberrantSplicing_pipeline_Counting_02_psi_value_calculation_FraseR_R

I tried using h5ls in R to identify the .o5_o3_psi5_psi3_dimnames/2 dataset that caused the error but could not source it in any of the .h5 files.

As a workaround I have had to delete all results and intermediate files and start from scratch to avoid the error, but this takes unecessary time and resources.

I am running DROP on a cluster and submitting each job independently.

DROP version 1.4.0

I have attached the full log file, my batch script, and cluster settings.

drop_cluster_fibro.sh.txt

26995003.drop_fibro.err.txt

config.yaml.txt

[AtaJadidAhari]

Hi @ChiaraF32,

Unfortunately, I cannot replicate your issue. However, when you add new samples, most of the steps generally have to be run again. By adding new samples, first the splitReads are counted for them. This results in new junctions which results in re-running nonSplitReads step for all samples, even the old ones. So I don't think any "unnecessary" jobs are being run in your case.

Hope this helps, let me know if anything is still unclear.

[akosycarz]

Hi, I get the same error when I want to rerun incomplete code or also when i wanted to run exportCounts. Did you manage to solve this probelm?

[vyepez88]

Hi @akosycarz , which FRASER version are you using? Can you use 2.4.3 (latest main in GitHub https://github.com/gagneurlab/fraser), that could resolve the issue.

[akosycarz]

ok, thank you