raredisease icon indicating copy to clipboard operation
raredisease copied to clipboard

Mismatch with new DNAScope ML model format

Open Jakob37 opened this issue 1 year ago • 3 comments

Description of the bug

The Sentieon DNAScope takes a flag --model which is expected to point to a file with a pre-trained model.

In their latest updates, the models are provided in .bundle format. Here the model is accessed as such:

--model /path/to/models.bundle/dnascope.model

(More info in issue here: https://github.com/Sentieon/sentieon-models/issues/4)

But, the file is actually named models.bundle.

So if you point to the models.bundle directly, the Sentieon process does currently not find the file. If pointing directly to dnascope.model, Nextflow will detect it as missing and will not run.

My temporary solution is to update the Sentieon DNA-scope module code as such:

def model_cmd                 = ml_model                           ? " --model ${ml_model}/dnascope.model"   : ''

Command used and terminal output

No response

Relevant files

No response

System information

No response

Jakob37 avatar Jun 13 '24 11:06 Jakob37

Yeah, we need to fix that. In the mean time, you can unpack the bundle using ar and get hold of the module-file you need.

EDIT: Your fix is also fine.

asp8200 avatar Jun 13 '24 12:06 asp8200

Yeah, we need to fix that. In the mean time, you can unpack the bundle using ar and get hold of the module-file you need.

EDIT: Your fix is also fine.

Didn't know that, thanks!

Jakob37 avatar Jun 13 '24 13:06 Jakob37

xref: https://github.com/nf-core/sarek/issues/1399

maxulysse avatar Aug 20 '24 09:08 maxulysse