Consider switching default runs to mosdepth - faster with similar resources

[sorelfitzgibbon]

mosdepth (Brent Pedersen, Aaron Quinlan), for depth of coverage calculations, supports threads (decompression threads with speed ups limited to 4 cpu). It is faster than CollectWgsMetrics and samtools depth and has a variety of useful options and outputs for WGS and targeted sequencing. It uses only around 1-2 GB of memory.

Comparison:

Coverage differences are likely due to different cutoffs that can be changed.

Tool	Sample	BAM size	run time	cpu	memory	mean coverage
CollectWgsMetrics	PCAWG-63_SA65868.bam	177 GB	4h 24m	54%	1.3 GB	29.4x
CollectWgsMetrics	PCAWG-63_SA65784.bam	295 GB	6h 42m	62%	1.3 GB	50.2x
mosdepth	PCAWG-63_SA65868.bam	177 GB	1h 23m	<300%	< 2 GB	34.5x
mosdepth	PCAWG-63_SA65784.bam	295 GB	2h 38m	<300%	< 2 GB	56x

It also has an option for d4 output format - "better than bigwig"

CollectWgsMetrics trace files:

/hot/user/sfitzgibbon/PCAWG-63/data/pipeline-runs/generate-SQC-BAM/generate-SQC-BAM-1.0.0/DO2629/log-generate-SQC-BAM-1.0.0-20240428T022515Z/nextflow-log/trace.txt

/hot/user/sfitzgibbon/PCAWG-63/data/pipeline-runs/generate-SQC-BAM/generate-SQC-BAM-1.0.0/DO2629/log-generate-SQC-BAM-1.0.0-20240430T211711Z/nextflow-log/trace.txt

CollectWgsMetrics output:

/hot/user/sfitzgibbon/PCAWG-63/data/pipeline-runs/generate-SQC-BAM/generate-SQC-BAM-1.0.0/DO2629/Picard-3.1.0/output/Picard-3.1.0_PCAWG-63_SA65784_wgs-metrics.txt

/hot/user/sfitzgibbon/PCAWG-63/data/pipeline-runs/generate-SQC-BAM/generate-SQC-BAM-1.0.0/DO2629/Picard-3.1.0/output/Picard-3.1.0_PCAWG-63_SA65868_wgs-metrics.txt

mosdepth output:

Windows of 500bp seem useful for WGS data. Overall mean coverage is at the bottom of the .summary.txt files.

/hot/data/unregistered/PCAWG-63/pipeline-runs/mosdepth/BWA-MEM2-2.2.1_GATK-4.2.4.1_PCAWG-63_SA65784.window500*

/hot/data/unregistered/PCAWG-63/pipeline-runs/mosdepth/BWA-MEM2-2.2.1_GATK-4.2.4.1_PCAWG-63_SA65784.window500*