dsl2 port

[phue]

This is some initial effort on porting nf-core/methylseq to dsl2.

My aim was to retain the functionality that was already there, I think that all features from v1.5 are working now.

A breaking change however is, that the pipeline now requires a samplesheet similar to what is already used in nf-core/nanoseq for example. It is supposed to have 4 columns:

sample	fastq_1	fastq_2	genome

The idea behind this change is to enable mapping of samples against different references (#181), something that is very useful for certain use cases.
Bonus: the samplesheet makes the single_end parameter obsolete

Would be great to get some opinions on this @nf-core/core

TODOs:

• update nf-core/test-datasets to reflect the change to samplesheet input methylseq: add samplesheet.csv test-datasets#214
• cleanup files that are not needed anymore:
  • Dockerfile, environment.yml
  • markdown_to_html.py (?)
• update github actions, we don't need docker builds anymore
• figure out why workflow summary has NAs
• add new modules to nf-core/modules modules for bisulfite sequencing data modules#129
  • bismark/[genome_preparation,align,deduplicate,extract,report,summary]
  • qualimap/bamqc
  • samtools/faidx
  • bwameth/align + bwameth/index
  • methyldackel/mbias + methyldackel/extract
• sync with latest version of nf-core modules
• create local modules where needed (especially with the feature outlined in Mapping samples to multiple reference fasta #181), this will be necessary for many of them
• figure out the right place to do bismark resource adjustments such as this one (should be passed via options.args to the module):

  methylseq/main.nf

  Lines 585 to 615 in 031be37

  	multicore = ''
  	if( task.cpus ){
  	// Numbers based on recommendation by Felix for a typical mouse genome
  	if( params.single_cell || params.zymo || params.non_directional ){
  	cpu_per_multicore = 5
  	mem_per_multicore = (18.GB).toBytes()
  	} else {
  	cpu_per_multicore = 3
  	mem_per_multicore = (13.GB).toBytes()
  	}
  	// Check if the user has specified this and overwrite if so
  	if(params.bismark_align_cpu_per_multicore) {
  	cpu_per_multicore = (params.bismark_align_cpu_per_multicore as int)
  	}
  	if(params.bismark_align_mem_per_multicore) {
  	mem_per_multicore = (params.bismark_align_mem_per_multicore as nextflow.util.MemoryUnit).toBytes()
  	}
  	// How many multicore splits can we afford with the cpus we have?
  	ccore = ((task.cpus as int) / cpu_per_multicore) as int
  	// Check that we have enough memory, assuming 13GB memory per instance (typical for mouse alignment)
  	try {
  	tmem = (task.memory as nextflow.util.MemoryUnit).toBytes()
  	mcore = (tmem / mem_per_multicore) as int
  	ccore = Math.min(ccore, mcore)
  	} catch (all) {
  	log.debug "Warning: Not able to define bismark align multicore based on available memory"
  	}
  	if( ccore > 1 ){
  	multicore = "--multicore $ccore"
  	}
  	}

• create a local MultiQC module
• test, test, test

[drpatelh]

Ah man! I will never get tired of seeing these initial DSL2 PRs appearing out of nowhere 😍 Looks great by just looking at the file changes!

The fact that nf-core/modules will get more and more padded out is a huge bonus too!

Nice work 🕺🏽

[phue]

@drpatelh Your nf-core/rnaseq port was a very helpful guideline to figure out how to do things! Thanks for that 👍

[phue]

closing this because there is now a dsl2 branch here