Good afternoon! Deepvariant (toolshed.g2.bx.psu.edu/repos/iuc/deepvariant/deepvariant/1.9.0+galaxy0) keeps terminating a task prematurely for the past two weeks. I tried to rerun the task that had been completed in the past and still got an error. According to the AI assistant, the error includes: oneDNN Custom Operations Notice, Missing CUDA Drivers, CPU Instructions Notice, Missing TensorRT Warning, and Re-use of Directory for Intermediate Results. I’d appreciate it if you could help troubleshoot the issue. For more information on the failed task, please refer to the most recent run with History API ID bbd44e69cb8906b58ddfa9fc977a29d4. Thank you!
Welcome @biology_tutorial
Hopefully we can help!
The uploaded BAM datasets uses chromosome identifiers like this (Ensembl format): 1, 2, ..
While the human reference genome you are using from UCSC uses chromosome identifiers like this: chr1, chr2,…
- FAQ: Mismatched Chromosome identifiers and how to avoid them
- More about how custom genomes work in Galaxy can be found in topics tagged with custom-genome
This tool wouldn’t have run in the past unless one of these was adjusted differently last time (whether in Galaxy or somewhere else – tools cannot do the mapping automatically usually). You could remap inside of Galaxy instead. The UCSC version of the hg38 reference genome is already indexed. Your alternative is to load your specific reference genome (what was used for mapping) up to Galaxy and to use that instead.
This guide covers more of these reference assembly version details. Even though is helping mostly with genomes native to the server the general advice will work for custom genome choices too! For variant analysis, you’ll want to pay attention to the labels and the bases between versions or the results will be incorrect due to the different base-coordinate systems.
Please let us know if this helps or if you have questions! 
Thanks a lot for the reviews and suggestions! I will try to correct the discrepancies and run DV again! I will keep you posted.
1 Like