Skip to content

wrap dataloader and model init in try block for MMSplice#91

Open
ramprasadn wants to merge 1 commit intokircherlab:masterfrom
ramprasadn:fixmmsplice
Open

wrap dataloader and model init in try block for MMSplice#91
ramprasadn wants to merge 1 commit intokircherlab:masterfrom
ramprasadn:fixmmsplice

Conversation

@ramprasadn
Copy link

Previously, only the predict_table_fast call was inside the try/except ValueError block. If SplicingVCFDataloader or MMSplice() raised a ValueError, which can happen when no input variants fall within splice regions, the exception was unhandled and the script would crash.

This fix extends the try block to include the dataloader and model initialization, ensuring the same graceful fallback to an empty pred_dict applies regardless of where the ValueError originates.

  Extends the ValueError catch to include SplicingVCFDataloader and
  MMSplice() initialization, not just the prediction step, to handle
  cases where no variants fall in splice regions.
@visze
Copy link
Collaborator

visze commented Mar 2, 2026

@ramprasadn Thank you! Can you give me an example variant to check?

@ramprasadn
Copy link
Author

Sure! Here is the file that caused MMsplice to crash.
cadd_mtinput.vcf.gz

@visze
Copy link
Collaborator

visze commented Mar 4, 2026

@ramprasadn I looked at your data. CADD does not support mitochondrial variants (chrM), and all of your variants are on chrM. So you won’t get any CADD scores from them anyway. It’s absolutely fine if the pipeline fails with this input. On our web services, we have a filter script upstream that removes such cases before they are processed by the CADD scripts.

Do you have any other variants on chr1–22 or chrX where you see the same issue?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants