Overview

The align-DNA Nextflow pipeline aligns paired-end data using BWA-MEM2, minibwa, and/or HISAT2, followed by processing with Picard Tools and SAMtools. The pipeline has been engineered to run in a 4 layer stack in a cloud-based scalable environment of CycleCloud, Slurm, Nextflow and Docker. Additionally, it has been validated with the SMC-HET dataset and reference GRCh38, where paired-end fastq's were created with BAM Surgeon.

The pipeline should be run WITH A SINGLE SAMPLE AT A TIME. Otherwise resource allocation and Nextflow errors could cause the pipeline to fail.

Developer's Notes:

For some reads with low mapping qualities, BWA-MEM2 assigns them to different genomic positions when using different CPU-numbers. If you want to 100% reproduce a run, the same CPU-number (bwa_mem_number_of_cpus) needs to be set.

BWA-MEM2 now only supports five CPU instruction set, AVX, AVX2, AVX512, SSE4.1 and SSE4.2. However we only tested the pipeline on AVX2 and AVX512 CPUs.

minibwa requires SSE4.2 on x86 CPUs or NEON on ARM CPUs. minibwa currently does not properly support alternate contigs; use a reference without alternate contigs.

We performed a benchmarking on our SLURM cluster. Using 56 CPUs for alignment (bwa_mem_number_of_cpus) gives it the best performance. See Testing and Validation.

---