GenMPI

MPI-based Cluster Scalable Variant Calling forShort/Long Reads Sequencing Data

GenMPI, an MPI based scalable method for both widely used short and long reads aligners, BWA-MEM and Minimap2, respectively. One of the main goals of GenMPI is to ensure 100% identical variant output compared to the single node baseline. In addition, GenMPI provides a flexible architecture which can be used to integrate a variety of alignment and variant calling tools.

Overall architecture:

Prerequisite:

Compiler

MPI runtime (impi/2021.2.0-intel-compilers-2021.2.0)

Make sure you have installed following tools:

seqkit
samtools
sambamba
vcftools

Aligners

bwa
minimap2

To use bwa-mem aligner, make sure you have above mentioned MPI complier installed:

cd bwa; make

To use minimap2 aligner, make sure you have above mentioned MPI complier installed:

cd minimap2; make

All other Variant Callers and phasing tools to be installed as required:

DeepVariant
Clair3
Octopus
GATK
Whatshap

Usage:

Note: We are using singilarity containers images for DeepVariant, Octopus, Clair3, Whatshap.

For short-reads

#split FASTQ
seqkit split2 --threads=128 -1 HG002.fastq -p 8 -O HG002/parts -f

#run MPI bwa aligner implementation on cluster
time srun -N 2 -n 2 minimap2 -t 128 -ax map-ont -R '@RG\tID:None\tSM:sample\tPL:Pacbio\tLB:sample\tPU:lane' reference/GRCh38_no_alt_analysis_set.fasta ONT/HG002.fastq > ONT/out.sam

#Compile the simple `queue_short.c` script which runs the whole variant calling worlflow
gcc queue_short.c -o queue_short

#run parallel MPI variant calling on cluster
time srun -N 1 -n 1 queue_short reference/GRCh38_no_alt_analysis_set.fasta ONT/output/ tools_path/ 1

For long-reads

#split FASTQ
seqkit split2 --threads=128 -1 HG002.fastq -p 8 -O HG002/parts -f

#run MPI minimap2 aligner implementation on cluster
time srun -N 2 -n 2 minimap2 -t 128 -ax map-ont -R '@RG\tID:None\tSM:sample\tPL:Pacbio\tLB:sample\tPU:lane' reference/GRCh38_no_alt_analysis_set.fasta ONT/HG002.fastq > ONT/out.sam

#Compile the simple `queue_long.c` script which runs the whole variant calling worlflow
gcc queue_long.c -o queue_long

#run parallel MPI variant calling on cluster
time srun -N 1 -n 1 queue_long reference/GRCh38_no_alt_analysis_set.fasta NT/output/ tools_path/ 1

Tanveer Ahmad et al., "GenMPI: Cluster Scalable Variant Calling for Short/Long Reads Sequencing Data", available at: biorxiv