Add modules outlined in the pipeline proposal

[kedhammar]

Functionalities and modules

Mentioned in the pipeline proposal

primaryQC_pipeline_proposal.pdf

Pipeline proposal Slack thread

Standard QC

• [x] FastQC Standard QC
• [ ] SeqKit histograms
  • I assume this refers to seqkit watch
  • Not available as a module currently, what I can see
• [ ] SeqFu
  • Not available as a module currently, what I can see
  • [ ] fastq integrity --> seqfu check
  • [ ] metadata --> seqfu metadata
  • [ ] merging lanes --> seqfu merge

Duplication + Complexity

• [ ] Preseq complexity
  • Which subtool?
• [ ] BBtools Clumpify
• [ ] UMI detection (stretch goal)

Adapter and Artifact detection

• [ ] Fastp

• [ ] BBtools

  • BBDuk
  • Testformat2
  • (RQCFilter2 is a corresponding subworkflow using multiple BBtools)
  • (For PacBio: Removesmartbell, Icecreamfinder)

  Contamination detection

• [ ] FastQ screen

• [ ] Sylph

• [ ] Kraken2

• [ ] Mapping to reference

Mentioned in the pipeline Slack channel

• [ ] Mash screen
• [ ] checkQC

[kedhammar]

6 PR draft to start adding modules

[mahesh-panchal]

For WGS data for assembly, GenomeScope (https://github.com/nf-core/modules/blob/master/modules/nf-core/genomescope2/main.nf). The database is built using Meryl ( also on nf-core ).

But there is also a container only version that's a little bit faster and has extra tools that might be useful (https://github.com/nf-core/modules/blob/master/modules/nf-core/genescopefk/main.nf) The databases for Merquryfk/KATGC, Merquryfk/KATCOMP, Merqury/Ploidyplot, and GeneScopefk are build using FastK.

[remiolsen]

Preseq complexity (which subtool?).

I've used preseq lc_extrap before and there's a module for it in nf-core (https://nf-co.re/modules/preseq_lcextrap). However, it is very prone to not working or rather refusing to give a complexity estimate.

Another option would be Picard (https://gatk.broadinstitute.org/hc/en-us/articles/360037591931-EstimateLibraryComplexity-Picard). I've never used it, and for the applications I worry about library complexity (HiC) the tool I use (pairtools) implemented it's own complexity estimate, so I have no need. There's no nf-core module for it as far as I can see.

[kedhammar]

Preseq complexity (which subtool?).

I've used preseq lc_extrap before and there's a module for it in nf-core (https://nf-co.re/modules/preseq_lcextrap). However, it is very prone to not working or rather refusing to give a complexity estimate.

Another option would be Picard (https://gatk.broadinstitute.org/hc/en-us/articles/360037591931-EstimateLibraryComplexity-Picard). I've never used it, and for the applications I worry about library complexity (HiC) the tool I use (pairtools) implemented it's own complexity estimate, so I have no need. There's no nf-core module for it as far as I can see.

@remiolsen any idea why preseq lc_extrap tends to refuse?

[remiolsen]

@remiolsen any idea why preseq lc_extrap tends to refuse?

I'm fairly certain I used to see this error most commonly - and I quote from the preseq manual

Q — When running lc extrap, I receive the error
ERROR: too many iterations, poor sample

A. — Most commonly this is due to the presence of defects in the approximation which cause the
estimates to be unstable. Setting the step size larger (with the flag -s) will help to avoid the
defects. The default step size is 1M reads or 0.05% of the input sample size rounded up to the
nearest million, whichever is larger. A consequence of this action will be a reduction in the
observed smoothness of the curve.

And setting the step -s flag was a little bit hit or miss if it worked.

[kedhammar]

Closed https://github.com/nf-core/seqinspector/pull/6 due to being too broad and unspecific. Feel free to start new PRs addressing more specific implementations.