Nextflow Workflow: Long-Read Somatic Variant Calling

Nextflow processes contain a collection of scripts that perform alignment, small-Variant calling, and haplotagging.

Currently pipeline includes the following:

• Alignment with minimap2
• Small variant calling with Clair3
• Phasing with longphase
• Somatic SV calling with Severus
• CNA calling with Wakhan
• Somatic small variant calling with DeepSomatic

Running on Biowulf/Slurm

Note that the commands below need to be run either in interactive session, or using sbatch, but not on the head node. The nextflow engine automatically submits jobs via Slurm scheduler.

Tumor-normal version

module load nextflow
nextflow run tumorNormalONT.nf --reads_tumor TUMOR_BAM --reads_normal NORMAL_BAM --reference REF_FASTA --outdir NF_OUT --vntr VNTR_BED --clair3_model CLAIR3_MODEL

Tumor-only version

module load nextflow
nextflow run ~/projects/LongReadCancerPipeline/tumorOnlyONT.nf --reads READS_BAM --reference REF_FASTA --outdir NF_OUT --vntr VNTR_BED --sv_pon SEVERUS_PON --clair3_model CLAIR3_MODEL

Some caveats

• Input reads are in (unmapped) bam format. You can specify multiple input files, but the whole argument needs to be wrapped in single quotes ('): --reads_tumor 'BAM1 BAM2', or reads_tumor 'BAM_DIR/*bam'
• If you are running multiple instances of nextflow (e.g. processing multiple samples), each instance needs to be run from separate working directory. This is because they create work folder for intermediate files and .nextflow logs.
• If pipeline failed for some reason it can be restarted by adding -resume argument (note single dash). You can modify the pipeline and restart, and it will attemnt to reuse the results if possible. Restart should happen inside the same working directory.

Resources to learn about Nextflow

• Highly suggest to go over this introduction if this is your first time with Nextflow: https://training.nextflow.io/

Tips for developing/debugging

• To run the pipeline locally, comment out process.executor = 'slurm' in the nextflow.config file
• Nextflow will report running/completed jobs as follows [11/4852c3] tumorOnlyOntWorkflow:alignMinimap2 (1) [ 0%] 0 of 1. You can find the associated files for this job at the (unique) working directory: work/11/4852c3....
• There, you'll see staged input files and everything that job output. There are also hidden files (start with ., use ls -la to show).
• In the working dir: .command.sh shows the executed command line (run inside of singularity container). .command.out - stdout, .command.err - stderr.
• Each process is run inside a singularity/docker container (referenced in the process definition). Build scripts for some of the custom docker containers are in docker folder.