ONT Oxford Nanopore fast5 processing

Software Requirements

an initial python venv

bonito has a very specific requirement on pytorch 1.10.0, so depending on the base CUDA version we'll look at setting up one of these two:

# CUDA 11.1

pip install torch==1.10.0+cu111 torchvision==0.11.0+cu111 torchaudio==0.10.0 -f https://download.pytorch.org/whl/torch_stable.html
# CUDA 10.2
pip install torch==1.10.0+cu102 torchvision==0.11.0+cu102 torchaudio==0.10.0 -f https://download.pytorch.org/whl/torch_stable.html

since we're wanting to use the A100's I'm assuming that they're running CUDA 11, but just wanted to capture this note

bonito

guppy (I see there is a nightly build of IGV that supports the new modbam tags, so worth testing)

samtools

bedtools

minimap2

nextflow

clair3 (as implemented below)

the specific workflow: nextflow run epi2me-labs/wf-human-snp --help

whatshap

cuteSV

sniffles2

IGV (though this might need some work as it's GUI based)

could look at the web server version

Bonito

eresearchqut/bonito: A docker image for the Oxford Nanopore Technologies Bonito software (github.com)

singularity exec -B /work/ont --nv docker://ghcr.io/eresearchqut/bonito:v0.0.3 bonito basecaller dna_r9.4.1_e8_sup@v3.3 \
    . \
    --modified-bases 5mC \
    --reference /work/ont/reference/GCA_000001405.15_GRCh38_no_alt_analysis_set.mmi \
    --recursive \
    --alignment-threads 4 > basecalls_mod_ref_S.bam

This example runs bonito from the container. It uses a particular folder with fast5 files (line 2)

Modified bases parameter 5mC (line 3)

Path to reference (line 4)

Search folder recursively (line 5)

Use threads and redirect STDOUT to a file (line 6)

Bonito Standalone

#!/bin/bash -l
#PBS -N bonito-gpu
#PBS -l select=1:ncpus=8:mem=32gb:ngpus=1:gputype=A100
#PBS -l walltime=72:00:00

#Change to folder where the job was submitted
cd $PBS_O_WORKDIR

# Run Bonito - check paths and parameters are correct
singularity exec -B /work/ont --nv docker://ghcr.io/eresearchqut/bonito:v0.0.5 bonito \
  basecaller dna_r9.4.1_e8_sup@v3.3 \
  ./example_fast5 \
  --modified-bases 5mC \
  --reference ./reference/GCA_000001405.15_GRCh38_no_alt_analysis_set.mmi \
  --recursive \
  --alignment-threads 4 | \
  samtools view -u | samtools sort -@ 4 > ./bonito_test/bam/HG002_as_test.bam
  
# Run samtools index
singularity exec -B /work/ont --nv docker://ghcr.io/eresearchqut/bonito:v0.0.5 \
  samtools index HG002_as_test.bam

Clair3

Clair3 is available in a nextflow pipeline:

epi2me-labs/wf-human-snp: Small variant calling for human samples (github.com)

The pipeline’s configuration is to run in the local process and not submit jobs - will need to test if the overhead of PBS compares with running “locally” in a job.

Using test data:

nextflow run epi2me-labs/wf-human-snp -profile singularity \
    --bam /work/ont/clair3/sample_data/chr6_chr20.bam \
    --bed /work/ont/clair3/sample_data/chr6_chr20.bed \
    --ref /work/ont/clair3/sample_data/chr6_chr20.fasta \
    --out_dir "results" \
    -process.executor "pbspro"

Line 1: Use singularity for the pipeline software

Line 5: save results in the results folder

Line 6: use PBS to run the processes