bohra 1.0.25

A bioinformatics pipeline for analysing short read Illumina data microbiological public health.

Bohra

Bohra is microbial genomics pipeline, designed predominantly for use in public health, but may also be useful in research settings. The pipeline takes as input a tab-delimited file with the isolate IDs followed by the path to READ1 and READ2, a reference for alignment and a unique identifier, where reads are illumina paired end reads (other platforms are not supported).

Motivation

Bohra was inspired by Nullarbor (https://github.com/tseemann/nullarbor) to be used in public health microbiology labs for analysis of short reads from microbiological samples. The pipeline is written in Snakemake.

Etymology

Bohra the name of an exinct species of tree kangaroo that lived on the nullarbor. The name was chosen to reflect the fact that it will be predominantly used to build trees, relies on snippy (named for a very famous kangaroo) and was inspired by nullarbor.

Pipeline

Bohra takes raw sequencing reads and produces a standalone html file for simple distribution of reports.

Bohra can be run in three modes

1. SNPs and Phylogeny

• Clean reads
• Call variants
• Generate a phylogenetic tree

2. SNPs, Phylogeny, Typing, Annotation and Species Identification (DEFAULT)

• Clean reads
• Call variants
• Generate a phylogenetic tree
• Assemble
• Species identification
• MLST
• Resistome
• Annotate

3. SNPs, Phylogeny, PanGenome and Typing and Species Identification

• Clean reads
• Call variants
• Generate a phylogenetic tree
• Assemble
• Species identification
• MLST
• Resistome
• Annotate
• Pan Genome

Installation

Bohra requires >=python3.7

Conda (Recomended)

Installing bohra with conda will ensure that all dependencies are present. See below for instructions on how to configure the databases for kraken2.

Set up conda - documentation for conda installation can be found here

conda config --add channels defaults
conda config --add channels bioconda
conda config --add channels conda-forge

It is recomended that you set up a bohra environment

conda create -n <bohra_env_name> bohra

To use bohra

conda activate <bohra_env_name>

PyPi

If installing with pip you will need to ensure other dependencies are also installed.

pip3 install bohra

• Snippy
• Shovill (skesa and spades.py)
• Roary
• Prokka
• kraken2
• abricate
• mlst
• iqtree
• seqtk
• snp-dists

IMPORTANT In addition to installing kraken ensure that you have a kraken2 database. Minikraken can obtained as follows

wget ftp://ftp.ccb.jhu.edu/pub/data/kraken2_dbs/minikraken2_v2_8GB_201904_UPDATE.tgz
tar -C $HOME -zxvf minikraken2_v2_8GB_201904_UPDATE.tgz

This will download and unzip the kraken2 DB. Other kraken2 DB are also available, you can find more information here

Once you have the DB downloaded you will have to create an environment variable called KRAKEN2_DEFAULT_DB. This can be done by adding the following to your $HOME/.bashrc

export KRAKEN_DEFAULT_DB=$HOME/minikraken2_v2_8GB_201904_UPDATE

Bohra can be run in two modes run for an initial analysis and rerun for a re-analysis. A .html report is generated allowing for the visualisation of tree and examination of the dataset to provide insights that may be useful in interpretation of the results.

Set up

Input file

The input file needs to be a tab-delimited file with three columns IsolateID, path to R1 and path to R2.

Isolate-ID    /path/to/reads/R1.fq.gz    /path/to/reads/R2.fq.gz

Reference

The choice of reference is important for the accuracy of SNP detection and therefore the investigation of genomic relatedness. Appropriate references should be chosen following the guidelines below.

1. A closed reference from the same ST (where applicable) or a gold-standard reference (as may be used in M. tuberculosis).
2. A pacbio or nanopore assembly from MDU that is of the same type as the query dataset
3. A high quality de novo assembly of either an isolate in the dataset or an isolate of the same ST or type.

Mask

Phage masking is important for to prevent the inflation of SNPs that can be introduced by horizontal transfer as opposed to vertical transfer. For closed genomes or those that are publicly available phaster-query.pl can be used to identify regions for masking. If a denovo assembly is used the website phaster.ca can be used. Regions for masking should be provided in .bed format.

Run

Minimal command

bohra run -r path/to/reference -i path/to/inputfile -j unique_id -m path/to/maskfile (optional)

usage: bohra run [-h] [--input_file INPUT_FILE] [--job_id JOB_ID]
                 [--reference REFERENCE] [--mask MASK]
                 [--pipeline {sa,s,a,all}]
                 [--assembler {shovill,skesa,spades}] [--cpus CPUS]
                 [--minaln MINALN] [--prefillpath PREFILLPATH] [--mdu MDU]
                 [--workdir WORKDIR] [--resources RESOURCES] [--force]
                 [--dryrun] [--gubbins]

optional arguments:
  -h, --help            show this help message and exit
  --input_file INPUT_FILE, -i INPUT_FILE
                        Input file = tab-delimited with 3 columns
                        <isolatename> <path_to_read1> <path_to_read2>
                        (default: )
  --job_id JOB_ID, -j JOB_ID
                        Job ID, will be the name of the output directory
                        (default: )
  --reference REFERENCE, -r REFERENCE
                        Path to reference (.gbk or .fa) (default: )
  --mask MASK, -m MASK  Path to mask file if used (.bed) (default: False)
  --pipeline {sa,s,a,all}, -p {sa,s,a,all}
                        The pipeline to run. SNPS ('s') will call SNPs and
                        generate phylogeny, ASSEMBLIES ('a') will generate
                        assemblies and perform mlst and species identification
                        using kraken2, SNPs and ASSEMBLIES ('sa' - default)
                        will perform SNPs and ASSEMBLIES. ALL ('all') will
                        perform SNPS, ASSEMBLIES and ROARY for pan-genome
                        analysis (default: sa)
  --assembler {shovill,skesa,spades}, -a {shovill,skesa,spades}
                        Assembler to use. (default: shovill)
  --cpus CPUS, -c CPUS  Number of CPU cores to run, will define how many rules
                        are run at a time (default: 36)
  --minaln MINALN, -ma MINALN
                        Minimum percent alignment (default: 0)
  --prefillpath PREFILLPATH, -pf PREFILLPATH
                        Path to existing assemblies - in the form
                        path_to_somewhere/isolatename/contigs.fa (default:
                        None)
  --mdu MDU             If running on MDU data (default: True)
  --workdir WORKDIR, -w WORKDIR
                        Working directory, default is current directory
                        (default: /home/khhor)
  --resources RESOURCES, -s RESOURCES
                        Directory where templates are stored (default:
                        /home/khhor/dev/bohra/bohra/templates)
  --force, -f           Add if you would like to force a complete restart of
                        the pipeline. All previous logs will be lost.
                        (default: False)
  --dryrun, -n          If you would like to see a dry run of commands to be
                        executed. (default: False)
  --gubbins, -g         If you would like to run gubbins. NOT IN USE YET -
                        PLEASE DO NOT USE (default: False)
  --cluster, -clst      If you are running Bohra on a cluster. Note if set you 
                        will need to provide a cluster.json file and a 
                        run_snakemake.sh, you can see examples on the 
                        documentation page.
  --json                Path to cluster.json - required if --cluster is set
  --run-snake           Path to run_snakemake.sh - required if --cluster is set

Rerun

A rerun may be performed if changes to the reference and/or mask file are needed. In addition, if isolates need to be removed or added to the analysis. The following behaviour on a rerun should be expected;

• New reference will result in calling of snps in all isolates of the analysis
• If the reference is unchanged SNPs will only be called on new isolates
• Determination of core alignment, distances and generation of trees will occur for every rerun

-r and -m are only required if these are to be different to the previous run. If not Bohra will detect and use the previous reference and mask files. Also changes to the isolates included should be made to the input file used in the original run. New isolates can be added to the bottom of the input file and prefixing an isolate with # will remove it from the analysis.

Minimal command

bohra rerun

                   [--workdir WORKDIR] [--resources RESOURCES] [--dryrun]
                   [--gubbins] [--keep]

optional arguments:
  -h, --help            show this help message and exit
  --reference REFERENCE, -r REFERENCE
                        Path to reference (.gbk or .fa) (default: )
  --mask MASK, -m MASK  Path to mask file if used (.bed) (default: )
  --cpus CPUS, -c CPUS  Number of CPU cores to run, will define how many rules
                        are run at a time (default: 36)
  --workdir WORKDIR, -w WORKDIR
                        Working directory, default is current directory
                        (default: /home/khhor)
  --resources RESOURCES, -s RESOURCES
                        Directory where templates are stored (default:
                        /home/khhor/dev/bohra/bohra/templates)
  --dryrun, -n          If you would like to see a dry run of commands to be
                        executed. (default: False)
  --gubbins, -g         If you would like to run gubbins. NOT IN USE YET -
                        PLEASE DO NOT USE (default: False)
  --keep, -k            Keep report from previous run (default: False)
  --cluster, -clst      If you are running Bohra on a cluster. Note if set you 
                        will need to provide a cluster.json file and a 
                        run_snakemake.sh, you can see examples on the 
                        documentation page.
  --json                Path to cluster.json - if not included will default
                        to version provided in previous run
  --queue               The queueing system in use - qsub or sbatch

Running Bohra in a HPC environment

Bohra can be run in a HPC environment (currently only sbatch and qsub are supported). To do this some knowledge and experience in such environments is assumed. You will need to provide a file called cluster.json. This file will contain rule specifc and default settings for running the pipeline. An template is shown below (it is recommended that you use this template, settings have been established using a slurm queueing system), in addition you can see further documentation here.

example command

bohra run -r <reference> -i <input_tab> -j <jobname> --cluster --json cluster.json --queue sbatch

bohra rerun --cluster --queue sbatch

cluster.json

{
    "__default__" :
    {
        "account" : "AccountName",
        "time" : "0-0:5:00",
        "cpus-per-task": "2",
        "partition" : "cloud",
        "mem" : "2G",
        "ntasks" : "4",
        "job" : "{rule}"
    },
    "snippy" :
    {

        "cpus-per-task" : "4",
        "time" : "0-0:5:00",
        "mem" : "8G" 
    },
    "assemble":
    {
        "cpus-per-task" : "4",
        "time" : "0-0:20:00",
        "mem" : "32G"

    },
    "kraken" :
    {
        "cpus-per-task" : "8",
        "time" : "0-0:20:00",
        "mem"  : "32G"
    },
    "run_iqtree_core" :
    {
        "time": "0-0:20:00"
    },
    "roary" : 
    {
        "cpus-per-task" : "36",
        "time" : "0-0:25:00",
        "mem": "8G"
    },
    "run_prokka" :
    {
        "cpus-per-task" : "8",
        "time" : "0-0:10:00"
    },
    "run_snippy_core" :
    {
        "time" : "0-0:15:00"
    }
}