Community feedback needed: Consider other tools

[maxulysse]

Issue by @MaxUlysse, moved from SciLifeLab#666

• [ ] ExpansionHunter for estimating repeat sizes
• [ ] QDNAseq CNVs for shallow WGS
• [x] CNVkit for CNVs
• [ ] deconstructSigs for mutational signatures
• [ ] ditto mutation-signatures
• [ ] SomaticSeq ensemble caller and machine learning
• [x] MSIsensor for replication slippage variants at microsatellite regions
• [ ] Oncotator for annotating human point mutations and indels
• [ ] MutSig for Mutation Significance checks
• [ ] TrackSig to infer mutational signatures in cancer over time
• [ ] DriverPower to discover potential coding and non-coding cancer driver elements from tumour whole-genome or whole-exome somatic mutations
• [ ] SVclone for inferring the cancer cell fraction of tumour structural variation from whole-genome sequencing data
• [ ] HsMetrics better QC
• [x] CNNScoreVariants Annotate a VCF with scores from a Convolutional Neural Network (CNN)
• [ ] SignatureAnalyzer yet an other mutation signature analyser from Broad
• [ ] PROSIC2 is a caller for somatic variants in tumor-normal sample pairs
• [ ] seqCAT Evaluation of SNV profiles
• [ ] varlociraptor A Rust library for variant calling using a latent variable model
• [x] GATK CNN Convolutional Neural Net to filter annotated variants
• [x] Google DeepVariant Deep neural network to call genetic variants from next-generation DNA sequencing data
• [ ] Octopus
• [x] mosdepth fast BAM/CRAM depth calculation for WGS, exome, or targeted sequencing
• [ ] rock circos plot out of manta
• [ ] DeepTrio
• [ ] SmuRF for Random Forest Ensembl Somatic Calling
• [x] PureCN for estimating purity/ploidy
• [ ] Lofreq a fast and sensitive variant-caller for inferring SNVs and indels from next-generation sequencing data. LoFreq is very sensitive; most notably, it is able to predict variants below the average base-call quality (i.e. sequencing error rate).
• [ ] DeepSomatic extension of deep learning-based variant caller DeepVariant
• [ ] MEDICC2 Minimum Event Distance for Intra-tumour Copy-number Comparisons
• [ ] DPClust Dirichlet Process based methods for subclonal reconstruction of tumours
• [ ] LINX annotation, interpretation and visualisation tool for structural variants
• [ ] FindDNAFusion a combinatorial pipeline for the detection of cancer-associated gene fusions in next-generation DNA sequencing data

[davidmasp]

Would something like somalier be interesting? just saw in twitter that now it can be integrated in MultiQC. I am not sure if there is a conda package though...

[maxulysse]

Nothing yet in Bioconda. What would be your use case with somalier?

[davidmasp]

Maybe not superuseful in terms of new features. I understand it identifies that tumor/ normal come from the same patient. I think it could be useful in some set ups and maybe also somatic calling for cell lines for instance which aren’t exactly a classical use case for a somatic vc pipeline. Indeed idk about the tool much though.

[mvelinder]

I'd throw my support in for DeepVariant integration under --tools

Is that still a priority? It seemed like it was being worked on back in #105

Thanks!

[mvelinder]

ExpansionHunter would also be a great addition under --tools

[maxulysse]

Hi @mvelinder both are tools that we indeed considered. At the time of #105 deepvariant seemed to big to be integrated into sarek. with DSL2, it would make sense to have it within sarek, so we currently have plan to have all the necessary modules made in nf-core/modules and then integrate into sarek.

[egenomics]

Hi,

For clinical environments it would be great to have a depth calculation tool. Mosdepth is a winner, as already suggested.

Another important tool would be a cnv caller. Manta is not a good tool for it, at least for gene panels. An easy and validated solution would be Exomedepth: https://pubmed.ncbi.nlm.nih.gov/32561899/

[abhi18av]

I'd like to suggest addition of DeepTrio which builds upon DeepVariant, aimed at variant calling of trios or duos.

[FriederikeHanssen]

Expansionhunter & HSMetrics have both been suggested by @marchoeppner https://nfcore.slack.com/archives/CGFUX04HZ/p1647518666493429?thread_ts=1647517444.673109&cid=CGFUX04HZ something for 3.1?

[gorgitko]

Hi, deepvariant has been already incorporated, but without deepTrio, right?

[FriederikeHanssen]

yes that is right

[gorgitko]

Would it be possible to include it in TODO (board)? :innocent: Thanks!

[tdanhorn]

I second the inclusion of Picard Tools CollectHsMetrics (for WES/targeted sequencing), and there is CollectWgsMetrics for WGS. Picard is already there, so all that is needed is an option to run it. I find HsMetrics very useful for assessing coverage (e.g. it can tell you if deeper sequencing might solve a coverage problem or not), and while I can run it manually, it would be much nicer to just have it included. In addition to the genome and the BAM, it needs intervals for targets (i.e. the existing --target_bed parameter) and for baits (would need an equivalent parameter for this). These tools also work nicely with MultiQC. Thank you!

[praveenraj2018]

It would be nice to have more SV callers in Sarek for better characterization, especially for tumor and normal samples. It is always a challenge to figure-out true/significant calls when there are complex rearrangements exist.

I suggest to include:

GRIDSS2 - Comprehensive characterization of SVs (https://genomebiology.biomedcentral.com/articles/10.1186/s13059-021-02423-x)

and then,

LINX - algorithm classifies somatic structural variation in tumors; It does chaining, clustering, and visualizations provide insights into complex rearrangements (https://www.sciencedirect.com/science/article/pii/S2666979X22000325)

It has been demonstrated in several papers recently in oncology:

https://www.nature.com/articles/s41698-021-00155-6 https://www.nature.com/articles/s41591-020-1072-4.pdf?proof=tr https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7545150/ https://www.biorxiv.org/content/10.1101/781013v1 https://www.frontiersin.org/articles/10.3389/fonc.2022.919118/full https://pubmed.ncbi.nlm.nih.gov/35047820/

[mvelinder]

octopus also looked very promising

https://github.com/luntergroup/octopus

https://www.nature.com/articles/s41587-021-00861-3

[maxulysse]

I was sure it was on our list already

[maxulysse]

@lbeltrame (cc @LMannarino @aldosr @rikizad) to add a tool I'd recommend to create a modules in the nf-core/modules repo following this documentation: https://nf-co.re/developers/modules#writing-a-new-module-reference Please don't hesitate to ping me (either here or on Slack if you need any pointers).

[lbeltrame]

Thanks, Maxime. We have experience writing nf-core modules (we have a handful we use internally, ranging from local modifications to entire pipelines), but I'll make sure to ping if need be (tests can be... problematic ;).

I'll file separate issues on nf-core/modules to track this.

[maxulysse]

I look forward to the opportunity to collaborate

[nevinwu]

Hi all, I'm currently working on a lofreq_somatic module to perform somatic variant calling on paired samples with the lofreq caller. Once it is finished, maybe it would be interesting adding it to sarek as another caller. Tumor-only modes of lofreq have already been implemented and are part of nf-core/modules (lofreq_call and lofreq_callparallel).

[Lumimar]

I am wondering whether there might be scope to extend sarek functionality to add:

• Gem mapper - https://www.nature.com/articles/nmeth.2221 (review states “The Genome Multitool (GEM) mapper rapidly and accurately provides all alignments of a read within a user-defined number of mismatches.”). Github: https://github.com/smarco/gem3-mapper

• SiNPle (https://www.mdpi.com/2073-4425/10/8/561) “SiNPle (Simplified Inference of Novel Polymorphisms from Large coveragE), a fast and effective software for variant calling. SiNPle is based on a simplified Bayesian approach to compute the posterior probability that a variant is not generated by sequencing errors or PCR artefacts.” Github: https://github.com/PaoloRibeca/SiNPle

Many thanks!

[maxulysse]

Consider adding Normalisation for VCFs before annotation

[SofiaMAhmed]

I would like Octopus to be added to Sarek if possible

[maxulysse]

I would like Octopus to be added to Sarek if possible

I considered that a while back, but the installation was not easy, it seems that it's available on bioconda nowadays, so I'll definitively have a look into it: https://bioconda.github.io/recipes/octopus/README.html

Do you know if the tool is still maintained, I don't see much activity on the github repo

[eolaniru]

would be nice to have vardictjava , a sensitive variant caller added to sarek

[bounlu]

Any fusion detection tool such as LINX or FindDNAFusion will be a great addition.