Best practice Wakhan

[ljwharbers]

Hi,

Thanks for developing and the continuous improvements on Wakhan!

We currently have Wakhan incorporated in our nf-core style Nextflow pipeline: https://github.com/IntGenomicsLab/lrsomatic which supports ONT/PacBio (optional Fiber-seq) data and both tumor-only or TN paired.

Now we saw that you have made some updates, resulting in two modules hapcorrect and cna. I also saw that you have developed a nextflow pipeline as well to process long-read WGS data and where, in short, you do:

1. Alignment
2. Germline variant calling
3. Wakhan hapcorrect
4. Severus
5. Wakhan cna

I was now wondering what the reasoning and added benefit for hapcorrect is and if you do recommend us to also change our pipeline to include this step.

Best,
Luuk

[mikolmogorov]

Hi Luuk,

Happy to hear that Wakhan is helpful for your workflow and thanks a lot for the feedback!

wakhan hapcorrect and wakhan cna are basically the two halfs of the original Wakhan workflow. We've been experimenting with running Severus between hapcorrect and cna for two main reasons. First, Severus breakpoint phasing will be consistent with the Wakhan coverage profile. And second, Wakhan can also use the Severus SV phasing to put additional constraints on the CNA solution (by knowing which haplotype has the CNA) - this is enabled by --use-sv-haplotypes option.

I think for the many applications the practical difference of this change may be minimal in terms of SV and CNA calls. What can change is phasing of those variants on the chromosome-scale level. We hope to explain this better in a preprint sometime within the next few weeks.

I'd say we are still not 100% sure we will eventually recommend separate hapcorrect + cna runs as the default option to run Wakhan as it complicates the workflow engineering a bit, and we have a couple ideas how to make it more elegant. You are definitely welcome to give it a try, but just to be aware that it might change again. We do hope to have a final stable version of Wakhan sometime in the next couple months.

[ljwharbers]

Hi Mikhail,

Thanks for your response!

Having this as separate modules indeed makes sense to improve the consistency of the phasing and improve this for SVs at chromosome scale.

Since you're not completely sure yet on the 'final' implementation of this, we'll stick with the current implementation in the pipeline. We can definitely adapt it in the future if the recommendation on your end changes.

Cheers,
Luuk