Deepvariant Version

[kokyriakidis]

@ewels
Hello,

1. Could you update the pipeline to 0.72 version?

2. If i run it locally on my pc, will it use my acceleration cores of my GPU?

[apeltzer]

1.) I guess @PhilPalmer might be able to do that...
2.) No, I guess this would require a special container with CUDA support being built in a custom way - could be interesting though to support this at some point...

[PhilPalmer]

Hi @kokyriakidis,

1. I have tried updating the pipeline. It wasn't working with the test data for some reason though. If you also want to try it out I pushed the changes to the dv-v0.72 branch. I believe you can run it with nextflow by adding -r dv-v0.72 to your nextflow run command.
2. Yes I think @apeltzer has covered this. There are no plans to implement this but I will take note and see if other people have the same issue. Thanks for the feedback.

Sorry I couldn't be of more help. Let me know if you have any more questions and what the result of using the dv-v0.72 branch is if you decide to test it.

Thanks

[kokyriakidis]

Maybe it’s time to update to v0.8 which includes singularity support:

“
google/deepvariant#132 (comment)
“

And since it’s easy now to run the GPU enabled model, it would be lifesaver to include GPU enabled support in this nextflow pipeline

[apeltzer]

Before that, I'd rather wait for some improvements on GPU support in nextflow itself, (x-ref: nextflow-io/nextflow#997).

As we don't use the main docker/singularity images provided by the Deepvariant team in this pipeline for various reasons / easier maintenance for example, it might also require some adjustments in the image to be able to use it for GPU computing stuff.

[kokyriakidis]

Any update about implementing new version of deepvariant?

[PhilPalmer]

Hi,

Sorry for the late response.

In theory, it should be as simple as just updating this one line.

However, when I tried updating the version to 0.72 the pipeline was failing with the testdata. I hope to get round to building a docker container with deepvaraint version 0.8 & testing it soon.

[kokyriakidis]

Ok then! When it is ready, let us know! :-) Στις Τρί, 2 Ιουλ 2019 στις 12:44 μμ ο χρήστης PhilPalmer < [email protected]> έγραψε:

…

Hi, Sorry for the late response. Still no update. In theory, it should be as simple as just updating this one line <https://github.com/nf-core/deepvariant/blob/master/environment.yml#L9>. However, when I tried updating the version to 0.72 the pipeline was failing with the testdata. I hope to get round to building a docker container with deepvaraint version 0.8 & testing it soon. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#19?email_source=notifications&email_token=AIBLFXL2ZHHHA7DMXSMGWZTP5MPNHA5CNFSM4GX6LWHKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZAWRGQ#issuecomment-507603098>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AIBLFXP4W5BS7LTGAGNZ5ITP5MPNHANCNFSM4GX6LWHA> .

[marchoeppner]

HI, just bumping this. Deepvariant is now up to v 0.9.0 - with apparently significant improvements all around. So would be nice to see that included. Thanks!

[PhilPalmer]

Thanks for your patience. I agree that the DeepVariant version definitely needs to be updated.

I'm going to an nf-core hackathon 4-6th March so it could be a good opportunity to updated it then. Would you be able to wait until then or do you need this sooner?

[marchoeppner]

No worries, middle of March sounds fine.

[kokyriakidis]

Support for the GPU accelerated deepvariant would also be a nice feature!

[marchoeppner]

Just a friendly bumb :)

[marchoeppner]

Feeling bad to bump this again, but it's been another 2.5 months and DeepVariant is now up to version 0.10.0.

[ewels]

Bumping is never a bad thing! @PhilPalmer - any updates? There's another nf-core hackathon coming up soon... 😉

[ewels]

Note that doing any updates now will involve a little bit of work, as the pipeline code hasn't been touched for 2 years so it'll need synchronisation with the latest nf-core template.

[PhilPalmer]

Really sorry for the delay, I am definitely deserving of more than a bump 😅

This is something I've been meaning to do for a long while but never have the chance. Yeah, I think the pipeline will require quite a bit of refactoring. As well as the template synchronisation @ewels mentioned, I believe DeepVariant has undergone pretty significant updates itself and so whole processes can probably be removed. (The DeepVariant developers have written their own wrapper for the make_examples , call_variants & postprocess_variants so I think all of these processes could be combined into one)

@marchoeppner do you have any deadlines for when you would need the pipeline to be updated?

[marchoeppner]

Hi, thanks for looking into it. No, I have no specific deadline - for the moment, I am converting our CRAMs to BAM and then run v1.0 of the pipeline. Not elegant, but it works.