nf-core/pacsomatic: comprehensive somatic analysis using Pacbio HiFi read data

[wzhang42]

Pipeline title/name

pacsomatic

Keywords

cancer genomics, somatic analysis, HiFi read, variant (SNV, SV, CNV) calling, mutation signature, CpG methylation, Tumor purity, HRD

What is it about?

nf-core/pacsomatic is a bioinformatics nextflow pipeline which utilize the paired tumor/normal HiFi read data for comprehensive PacBio long-read somatic analysis. It includes a bunch of key functionalities such as HiFi read alignment, BAM coverage and quality control (QC), somatic SNV/SV/CNV detection and annotation, tumor purity/ploidy inference, homologous recombination deficiency (HRD) estimation, mutation signature analysis, CpG methylation profiling, and identification of differentially methylated regions (DMRs).
Key features:
1.) Adopted Nextflow DSL2 as the framework and singularity containers to run each functional module.
2.) A simple sample sheet as the sole input.
3.) Automatically pair tumor and normal BAM/VCF files for analysis.
4.) Support customizable run of the entire or part of pipeline.

Please provide a schematic diagram of the proposed pipeline

https://github.com/wzhang42/HiFi-Somatic-Nextflow/blob/main/Diagram.png

I confirm my proposed pipeline will follow nf-core guidelines. Most importantly, my pipeline will:

• be built with Nextflow.
• pass nf-core lint tests and use standardized parameters.
• be community-owned and developed within the nf-core organization.
• open source under the MIT license with proper credits and acknowledgments.
• have a descriptive, all lowercase, and without punctuation name.
• use the nf-core pipeline template and predominantly use official nf-core modules.
• focus on a specific data/analysis type with appropriate scope.
• have properly maintained documentation.
• be bundled using versioned Docker/Singularity containers.

Why do we need a new pipeline?

PacBio long-read sequencing, a third-generation sequencing (TGS) technology, offers several advantages over short-read sequencing. Somatic analysis are highly demanded in cancer/tumor studies. Currently, there is no Nextflow pipeline to use Pacbio HiFi reads for a comprehensive somatic analysis.

Who would be interested?

Tumor /cancer researchers.
Bioinformatics scientist in the field of Pacbio long read sequencing data analysis.

What has been done so far

We have developed most of the nextflow DSL2 codes. The initial test in HPC lsf cluster run well. I need more people involve and improve .

URL to existing work (if applicable)

https://github.com/wzhang42/HiFi-Somatic-Nextflow

Are there any similar existing nf-core pipelines?

none

[mashehu]

what's the overlap with genomeassembler? Any thought about this pipeline @nschan?

[nschan]

I think overlap with genomeassembler is minimal. As I understand this pipeline is for variant calling, not for assembly, going more in the direction of what sarek does?

[kenibrewer]

/approve

I think this is a valuable addition to the nf-core portfolio. It has a similar purpose to nf-core/sarek, but that pipeline already covers too many use cases. I like the idea of having this focused specifically on long-read data and specifically on tumor/normal somatic.

One question: Is PacBio long-read data different enough that it makes sense to keep it separate from other long-read platforms? Or would it potentially make sense to have this be longreadsomatic?

[famosab]

A couple of questions:

1. Are the names on the arrows the tool names you want to use? (for example somatic SNV calling with deepsomatic and clair3)?
2. How different are BAM files created from long-reads to BAM files created from short reads? Because there is a nf-core pipeline that is tailored for pacbio longread processing: https://nf-co.re/pacvar/dev/ (@tanyasarkjain knows more about that probably). So at least in the alignment step and the calling there might be considerable amount of overlap.

From my point of view merging your effort with nf-core/pacvar seems more sensible than a new pipeline :) But that would certainly help with the development of pacvar if more people work on that which would be great!

[wzhang42]

I am trying to answer all of the questions that you raised.
1.) This pipeline indeed is a somatic (tumor/normal pair) analysis, not a genome assembly pipeline. We are trying to cover the tumor somatic analysis tasks, such as somatic variant (SNV/SV/CNV) calling and annotation, mutation signature, CpG methylation/ DMR (Differential methylation region) detection between tumor and normal samples, and tumor purity inference.

2.) Some functional modules (eg, pb_pg_score, ) and the data format -Pacbio unaligned BAM (uBAM) are Pacbio-long read specific . Another long read platform ( eg. ONT) do not applied to. So, I feel more confident and prefer to give it a name as Pacbiosomatic or Pacbio-Hifi-somatic, not a more general longreadsomatic related.

3.) The tools in the arrow can be replaced. In the current version, we just devloped Deepsomatic and Clair3 for Pacbio somatic SNV calling. You know, nextflow framework is open and inclusive, which can easily integrate new tools with similar functions, or replace stale one.

4.). The BAMs in this diagram include two categories, unaligned BAM (uBAM, raw Pacbio sequencing reads), and alignment BAM ( after pbmm2 alignment). Some special tags are used in the BAMs. You can use higher-version samtools (samtools/17.0) to view them.

5.) We also developed our own Pacbio germine WGS analysis ( mainly germline variant calling ) nextflow pipeline. After check, I believe that our Pacbio WGS/germline nextflow pipeline is close to you mentioned pacbio longread processing pipeline -pacvar.
Somatic analysis is another very important and also a little bit complicate topic, which is fundamentally different to germline analysis, and suitable for cancer genomics studies, particularly for tumor/normal pair analysis.

[wzhang42]

@mashehu , I think , nschan has well addressed your questions. This pipeline we proposed is not about genome-assembly, but a pacbio long read somatic variant calling. @nschan , thanks for your answer

[wzhang42]

@kenibrewer many thanks for your positive comments about our pipeline and it's value in long-read tumor/normal somatic. To your suggestion to give it a broader name as longread-somatic to cover ONT platform, that's indeed a good idea, but several functional modules , e.g pb-cpg-score, and the unaligned Pacbio bam format are Pacbio-platform specific, ONT do not have the similar peer. I prefer to give it a name as pacsomtaic , just as another pacbio specific germline pipeline-paccar (https://github.com/nf-core/pacvar) .

[wzhang42]

@famosab, 1.) Of course, the tools in the arrow can be replaced. In the current version of this pipeline, Deepsomatic and Clair3 are ready and can be customized for Pacbio somatic SNV calling. Nextflow is an open and inclusive framework, we can easily integrate new tools with similar functions, or replace stale one. Once approved, we plan to investigate and add new tools. 2.) The BAMs in this diagram include two categories, unaligned BAM (uBAM, raw Pacbio sequencing reads), and alignment BAM ( after pbmm2 alignment). Some special tags are used in the BAMs. You can use higher-version samtools (samtools/17.0) to check them. 3.) I think, pacvar (https://nf-co.re/pacvar/dev/) is for pacbio germline analysis, our pipeline is for pacbio somatic analysis using tumor/normal pair. The alignment part (PBMM2) is the basic operation for pacvar and pacsomatic, but the downstream variant analysis are fundamentally different. Hope that I have addressed your questions.

[wzhang42]

Sorry that I just realized that I should specifically address your question one by one. Many thanks for your valuable questions.

[kenibrewer]

Having seen the pain points in nf-core/sarek from merging both germline and somatic analysis into a single pipeline, I would encourage this pipeline to remain separate from pacvar.

However, it would be good for the two pipelines to collaborate by creating shared nf-core/subworkflows were possible.

[wzhang42]

@kenibrewer, I totally agree . In fact, we also have another in-house nextflow pipeline for Pacbio germline analysis, which is close to pacvar. The alignment and other functional module that do not distinguish tumor and normal can be shared, but in most cases, the somatic analysis specifically strengthen the concepts of tumor/normal pair or tumor only, so they can not be directly apply to germline case.

Absolutely, the collaboration of the two pipelines will greatly benefit to pacvar and pacsomatic.

[jfy133]

So to summarise:

• This is basically a pacbio equivalent of sarek (typically Illumina)
• It distinguishes from pacvar because (presumably) of germline vs somatic?

I think this would render it sufficiently clear in scope that it does not overlap with the purpose of other pipelines?

That said I agree with @famosab that we should have @tanyasarkjain confirm that the somatic analyses would be out of scope for pacvar first.I do not see any explicit reference to germline or somatic in the pacvar repo nor paper.

[jfy133]

Also to through in an outsider but who I know works on PacBio @prototaxites any thoughts (even if outside of your normal area)