ellisrichardj / bovtb-nf

Nextflow script for processing WGS data

BovTB-nf

---

This is the updated pipeline for APHA's processing of Mycobacterium bovis WGS data. BovTB-nf is designed to process a batch (1 or more samples) of paired-end fastq files generated on an Illumina sequencer. It will first remove duplicate reads from the dataset (FastUniq) and then trim the unique reads based on base-call quality and the presence of adapters (Trimmomatic). Reads are then mapped to the M. bovis AF2122 reference genome and variants called (bwa/samtools/bcftools).

It has been built to run using nextflow, using standard bioinformatic tools for the most part. The external dependancies are:

• FastUniq
• Trimmomatic
• bwa
• samtools and bcftools
• bedtools
• Kraken2 (and database)
• Bracken

Installation

Of course Nextflow itself is a prerequisite and should be installed as described in the Nextflow Documentation

If you have the dependancies installed the pipeline can run by simply typing:

nextflow run ellisrichardj/BovTB-nf

Alternatively, clone the repository:

git clone https://github.com/ellisrichardj/BovTB-nf.git

If required, there is simple script for installing the dependancies (helpfully called Install_dependancies.sh), which will also update the nextflow config file with their locations.

Docker build

Alternatively, the pipeline can run in an ubuntu image on docker. To build the image: docker build /PATH/TO/REPO/ -t bov-tb

Run a docker container in bash: docker run --rm -it bov-tb

---

Examples

In its simplest form just run the Nextflow process from the directory containing the fastq files:

cd /path/to/Data
nextflow run BovTB-nf

bovtb-nf's Issues

Add explanations to output for failure to assign WGS Cluster

Currently if WGS cluster is not assigned, either bTB or LowCoverage is output. By using the data captured on %mapped and number of reads, it should be possible to determine if the sample should be repeated.

Output table of snps for each sample

A snp table similar to those produced by snippy would be a useful additional output.

Error in SNPfiltAnnot process

Updated reference genbank file leads to error in SNPfiltAnnot process:
AttributeError: 'CompoundLocation' object has no attribute '_start'

Reduce failures for contaminated samples

Samples are not assigned to WGS clusters when cultures are contaminated. Can kraken be used to filter out contaminants before mapping/cluster assignment?

Update trimmomatic version

Use latest version of trimmomatic

Bcftools consensus outputs reference base when data is missing

Need to change consensus calling so that when there is no data the reference sequence is not the output generated by default

Add run name to allow differentiation of the same sample sequenced multiple times

Collect folder name from $PWD?

VCF2Consensus process fails for some samples

Error code suggests that its linked to an issue in bcftools where two variants are overlapping. See samtools/bcftools#888. Fix available in latest bcftools commit, but not yet in a full release.

Update pipeline to use most recent version of M. bovis reference genome (LT708304)

Genotype inference

This part of the process needs work, both on the nextflow pipeline and the underlying python script to run correctly

Reliably detect M.bovis when it's at low prevalence in the sample

Need to filter kraken/bracken output so that presence of M.bovis in the sample is recognised and flagged. This should be added to the output csv file

Python script read stdin

The python script for snp filtering needs to be changed to accept stdin, so that the uncompressed vcf can be piped in rather than written to disk

Update Readme

Add some more useful information!

Update spoligotyping module to use standard names for spoligotypes

The internationally recognized names for spoligotypes are available here: https://www.mbovis.org/

Using --lowmem option causes kraken2 to fail

Infer phlogenetic cluster in addition to genotype

This will first require clusters to be defined, so is a long term goal.
Perhaps this can be done by identification of closest SNP pattern?

Formatting of BovPos.tsv output needs correcting

Some spaces need to be converted to tabs so that columns are displayed correctly

Add process to ID samples that are not M. bovis

Suggest run kraken on samples that do not give expected result