bronko is a viral variant caller that can rapidly detect most major and minor variants in any given sequencing dataset. bronko bypasses read mapping and variant calling and directly outputs a VCF from a set of reads and reference genome(s).
bronko also allows users to run multiple samples of the same species and build a multiple sequence alignment that can be inputted directly into phylogenetics software
- Simplicity -- bronko bypasses most steps of typical variant calling (indexing, read mapping, sam manipulation, variant calling, alignment, etc) and packages everything into 1-2 commands
- Large Scale Capability -- bronko allows you to test hundreds of strains against hundreds of samples simultaneously, allowing you to variant call diverse samples with diverse reference genomes
- Consistently ultrafast -- bronko is between 10-1000x faster than existing pipelines for read mapping (bowtie2, ie) and then variant calling (lofreq, ivar, ie), depending on the sequencing depth / number of reads as well as the number of threads available
- Comparable precision and sensitivity -- On our benchmarks so far, bronko achieves reasonable consistency with both ivar and lofreq on variant calling for both SNPs and iSNVs, even outperforming in some cases.
- If you are interested in identifying novel indels in viral genomes, currently we do not report on indel presence, but we are working on this problem. For historical sequencing data, we attempt to bypass this problem by letting you incorporate multiple reference genomes into a database and then we will automatically select the one with highest identity to each sample.
byonko bypasses readmapping by directly mapping kmers with small edit distance to a pileup representing the forward and reverse strands. It then uses the depth information and number of kmers mapping to each position/base to perform variant calling, similarly to existing tools. The lack of a formal pileup makes the process much more efficient, and thus we are able to achieve similar results in a fraction of the time.
Please perform quality control on your samples before running through this tool. In particular, please remove any primer sequences that were used, as this can have a large downstream impact on variant calls (for any software, but particularly using a kmer-based approach). It is also helpful to have reasonable base quality thresholds (>25 or >30) as bronko is unable to take this into account due to the transformation into kmer space.
For iSNVs, it is your responsibility to interpret the calls. Calling of iSNVs is known to be subject to a number of biases in the library preperation, PCR, and sequencing processes. We cannot control the input into bronko, and while we have found bronko to be very precise in its calls, it is intended to be used as a tool to identify putative iSNVs rather than declaring biological relevance.
The only non-rust requirement is KMC3 (https://github.com/refresh-bio/KMC), so please follow the instructions on their github to download the software on your system (either through conda or by downloading the source code directly). If you are downloading bronko using conda (recommended), you do not need to worry about this.
The best and recommended way to use bronko is through mamba/conda and the bioconda channel. If you do not have these, please follow this link to install them through miniforge: https://github.com/conda-forge/miniforge
Once installed, you should be able to download bronko using the following command
conda create -n bronko bronko
conda activate bronko
You should be able to test if installed successfully by running
bronko --help
For full details, please see the wiki.
bronko enables users to build indexes containing many strains of a single species. This can be used to run against a heterogenous dataset and the best reference will be chosen for each sample. If you have a single reference that you know you want to use, that is okay as well.
Basic usage is as follows:
bronko build -g /path/to/reference_genomes/*.fa -o bronko_db
This will output a bronko_db.bkdb index that you can use for downstream variant calling. The only parameter to change how this function is run is the kmer size k.
We are working on building a number of prebuilt databases for viruses of interest, so once these are available we will release them here. We also provide additional recommendations on how to build a bronko database to achieve the best results in the wiki.
To perform mapping / reference selection / variant calling, use the bronko call command:
If using a prebuilt database, use the -d/--db command
bronko call -d /path/to/bronko_db.bkdb -r /path/to/single_end_reads.fastq -o /path/to/output
or similarly, on paired end reads:
bronko call -d /path/to/bronko_db.bkdb -1 /path/to/reads/*_R1.fastq -2 /path/to/reads/*_R2.fastq -o /path/to/output
or both single end and paired end reads:
bronko call -d /path/to/bronko_db.bkdb -r /path/to/single_end_reads.fastq -1 /path/to/reads/*_R1.fastq -2 /path/to/reads/*_R2.fastq -o /path/to/output
Note: Keep single end reads in the -r and paired in the -1/-2. If the number of paired end reads do not match then an error will be thrown.
If you do not want to use a pre-built database, you can replace -d with -g and provide genomes in fasta format. This will still build a database, but not save it. If you are using a single genome, this may be sufficient.
There are several parameters that can be adjusted to fine-tune the stringency of variant calling. Refer to the wiki here for a comprehensive description of each parameter and its impact on downstream results when manipulated.
bronko is still in early versions and we welcome any feedback, suggestions, or feature requests. If you run into any issues, please email Ryan Doughty at [email protected] or raise an issue on github