1
0
Fork 0
mirror of https://github.com/MillironX/taxprofiler.git synced 2024-11-10 23:23:09 +00:00

Make adapter and quality trimming optional

This commit is contained in:
ljmesi 2022-03-21 18:17:08 +01:00
parent 24a01529f5
commit c97de32434
2 changed files with 10 additions and 3 deletions

View file

@ -57,6 +57,7 @@ params {
// FASTQ preprocessing
fastp_clip_merge = false
fastp_exclude_unmerged = true
remove_adapters = false
// MALT
run_malt = false

View file

@ -104,9 +104,16 @@ workflow TAXPROFILER {
FASTQ_PREPROCESSING ( INPUT_CHECK.out.fastq )
}
LONGREAD_PREPROCESSING ( INPUT_CHECK.out.nanopore )
ch_multiqc_files = Channel.empty()
if ( params.remove_adapters ) {
ch_longreads_preprocessed = LONGREAD_PREPROCESSING ( INPUT_CHECK.out.nanopore ).reads
.map { it -> [ it[0], [it[1]] ] }
ch_versions = ch_versions.mix(LONGREAD_PREPROCESSING.out.versions.first())
ch_multiqc_files = ch_multiqc_files.mix(LONGREAD_PREPROCESSING.out.mqc)
} else {
ch_longreads_preprocessed = INPUT_CHECK.out.nanopore
}
//
// PERFORM RUN MERGING
@ -138,6 +145,7 @@ workflow TAXPROFILER {
// output [DUMP: reads_plus_db] [['id':'2612', 'run_accession':'combined', 'instrument_platform':'ILLUMINA', 'single_end':1], <reads_path>/2612.merged.fastq.gz, ['tool':'malt', 'db_name':'mal95', 'db_params':'"-id 90"'], <db_path>/malt90]
ch_input_for_profiling = ch_reads_for_profiling
.mix( ch_longreads_preprocessed )
.combine(DB_CHECK.out.dbs)
.dump(tag: "reads_plus_db")
.branch {
@ -189,13 +197,11 @@ workflow TAXPROFILER {
workflow_summary = WorkflowTaxprofiler.paramsSummaryMultiqc(workflow, summary_params)
ch_workflow_summary = Channel.value(workflow_summary)
ch_multiqc_files = Channel.empty()
ch_multiqc_files = ch_multiqc_files.mix(Channel.from(ch_multiqc_config))
ch_multiqc_files = ch_multiqc_files.mix(ch_multiqc_custom_config.collect().ifEmpty([]))
ch_multiqc_files = ch_multiqc_files.mix(ch_workflow_summary.collectFile(name: 'workflow_summary_mqc.yaml'))
ch_multiqc_files = ch_multiqc_files.mix(CUSTOM_DUMPSOFTWAREVERSIONS.out.mqc_yml.collect())
ch_multiqc_files = ch_multiqc_files.mix(FASTQC.out.zip.collect{it[1]}.ifEmpty([]))
ch_multiqc_files = ch_multiqc_files.mix(LONGREAD_PREPROCESSING.out.mqc)
if (params.fastp_clip_merge) {
ch_multiqc_files = ch_multiqc_files.mix(FASTQ_PREPROCESSING.out.mqc)
}