Easier / lazy way to convert an existing bash pipeline to nextflow pipeline?

Question

Not going to lie, my head hurts when I read the NextFlow documentation (No dig on the documentation, I think my head is losing its processing capacity lol).

I had written a pipeline in Bash, essentially iterates over a text file containing full paths of R1 and R2 reads, followed by a series of commands and liberal use of gnu parallel to improve efficiency.

Is there any template I can follow, where I can slot in my commands and get things running quickly?

EDIT - The skeleton of pipeline goes as follows (As I'm not sure if I can put our inhouse pipeline in its entirety here) -

Bash code which loops through R1 and R2, aligns with BWA-mem2 and sorts the bam, then using gnu parallel, I essentially follow the steps as outlined in GATK Best Practices (looking to drop BQSR, but the steps remain) to obtain a gvcf file. The steps in Best Practices have been turned to functions, rather than each step being called sequentially.

Answer 1

If your objective is to convert Bash scripts into Nextflow code, you might find Viash (https://viash.io) a particularly useful open source tool. It automatically generates independent Nextflow modules from each of your scripts, along with the corresponding Docker containers for each module. Docker containers are generated on the fly based on your config description in Viash. These modules can be coupled together to build a (modular) Nextflow pipeline.

Answer 2

The gist would be something like:

fastqs = Channel.of(...some annoying groovy code to make this a list of tuples...)
workflow{
  align(fastqs)
  makeGVCF(align.out.bams)
  ... collect() the output from above joint call them all...
}

process align {
  input:
    tuple val(sampleName), path(reads)
  output:
    tuple val(sampleName), path("${sampleName}.bam"), emit: bams
  """
  bwa ... | samtools sort ...
  """
}

process makeGVCF{
  input:
    tuple val(sampleName), path(bam)
  output:
    path "${sampleName}.gvcf", emit: gvcfs
  """
  ...haplotype caller code...
  """
}

Answer 3

Since you asked about templates, if your task is something rather generic, there is a good chance that a module, or several, that you can use to build your workflow exists in https://nf-co.re/modules Generally I would recommend taking a look at the source of nf-core pipelines (e.g. https://nf-co.re/sarek for variant calling) to get an idea how to get your samplesheets in.