BACimpute

cgMLST Based Imputation for Bacterial Genes

BACimpute predicts missing allele calls in bacterial core-genome multilocus sequence typing (cgMLST) profiles. It combines chewBBACA schema and allele-calling workflows with statistical models trained on a reference collection of bacterial genomes.

! IMPORTANT ! BACimpute is under active development and is not yet a production release.

• How it works
• Getting Started
  • Quick Installation
  • Get / Update BACimpute
  • Call help
• Running BACimpute
  • Get a cgMLST Scheme
    • Download from Chewie-NS
    • Import from Ridom or Enterobase
    • Create a new Schema
  • Get an new Imputation Model
    • Test your new Model
• Parameter List
• Citations
• Contact

How it works

The workflow has the following main functionalities:

1. Obtain a cgMLST schema by downloading one from Chewie-NS, adapting an external schema, creating a schema from a reference genome collection, or supplying an existing chewBBACA schema.
2. Train and optionally evaluate an imputation model from the resulting allelic profiles.
3. Call alleles for new assemblies and impute loci reported as missing.

BACimpute provides two models:

• Inhomogeneous Markov chain (IMC): uses the physical order of loci in a reference genome. A missing locus is inferred from neighbouring loci; consecutive missing loci are handled with a forward-backward procedure.
• Maximum spanning tree (MaxST): connects loci according to normalized mutual information. Missing values are inferred by passing information between parent and child loci in the rooted tree.

Before model fitting, allele identifiers are converted to frequency ranks learned from the reference collection. Alleles not observed during training are represented as new alleles, while chewBBACA LNF calls are treated as missing values.

Getting Started

Quick Installation

To run the pipeline, you need to have Nextflow and either conda, Docker or Singularity.

Click! If you want to install Nextflow directly, you can use the following one-liner.

wget -qO- https://get.nextflow.io | bash

Click! If you want to set up conda to run the pipeline and install all other dependencies through it, you can use the following steps.

Use the following bash commands if you are working on Linux:

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh

Use the following bash commands if you are working on Mac:

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.sh
bash Miniconda3-latest-MacOSX-arm64.sh

Then, Nextflow an be installed over conda:

conda create -n nextflow -c bioconda nextflow
conda activate nextflow

Get / Update BACimpute

nextflow pull chkirschbaum/bacimpute

Call help

nextflow run chkirschbaum/bacimpute -r <version> --help

Running BACimpute

BACimpute takes as input either a directory with fasta sequences or a results_alleles.tsv from chewBBACA AlleleCall.

nextflow run chkirschbaum/bacimpute -r <version> -profile conda,local \
    --input 'profiles.tsv'

Get a cgMLST Scheme

Download from Chewie-NS

nextflow run chkirschbaum/bacimpute -r <version> -profile conda,local \
    --input 'profiles.tsv' \
    --new_schema \
    --schema_source 'chewie-ns' \
    --species 'Listeria monocytogenes' \
    --model_file 'path/to/model'

Click! If you want to see which schemes you can download from Chewie-NS.

The following schemes are available from Chewie-NS:

Species	Species ID (-sp)	Schema ID (-sc)
Streptococcus pyogenes	1	1
Acinobacter baumanii	2	1
Aliarcobacter butzleri	3	1
Bacillus anthracis	4	1
Brucella	5	1
Campylobacter jejuni	6	1
Clostridium chauvoei	7	1
Clostridium neonatale	8	1
Clostridium perfringens	9	1
Escherichia coli	10	1
Klebsiella oxytoca	11	1
Legionella longbeachae	12	1
Neisseria meningitidis	13	1
Salmonella enterica	14	1
Shewanella	15	1
Yersinia enterocolitica	16	1
Chlamydia trachomatis	17	1
Listeria monocytogenes	18	1
Legionella pneumophila	19	1
Yersinia pestis	20	1

Import from Ridom or Enterobase

nextflow run chkirschbaum/bacimpute -r <version> -profile conda,local \
    --input 'profiles.tsv' \
    --new_schema \
    --schema_source 'external' \
    --ext_schema_dir 'path/to/external_schema' \
    --model_file 'path/to/model'

Create a new Schema

Creating a schema requires either:

• a Prodigal training file ending in .trn; or
• a representative genome (GCF from NCBI RefSeq) ending in .fna, from which BACimpute generates a training file.

Several training files are provided in data/prodigal_training_files or via chewBBACA.

nextflow run chkirschbaum/bacimpute -r <version> -profile conda,local \
    --input 'profiles.tsv' \
    --new_schema \
    --schema_source 'create' \
    --ref_dir 'path/to/reference_set' \
    --training_file 'path/to/training_or_GCF_file' \
    --model_file 'path/to/model'

Get an new Imputation Model

nextflow run chkirschbaum/bacimpute -r <version> -profile conda,local \
    --input 'profiles.tsv' \
    --schema_dir 'path/to/schema' \
    --model 'imc' \
    --new_model \
    --ref 'refseq_name'

Test your new Model

You can test your new built model on the reference set you build the model on and a test set and get an interactive HTML report with the results.

nextflow run chkirschbaum/bacimpute -r <version> -profile conda,local \
    --input 'profiles.tsv' \
    --schema_dir 'path/to/schema' \
    --model 'imc' \
    --new_model \
    --ref 'refseq_name' \
    --ref_dir 'path/to/reference_set' \
    --test_dir 'path/to/test_set'

Parameter List

Parameter	Default	Description
--input	''	FASTA assembly or quoted glob of assemblies to impute.
--mode	clonal	Species handling: clonal or variable.
--schema_dir	''	Existing chewBBACA schema directory. Used when --new_schema is false.
--new_schema	false	Obtain or create a schema during the run.
--schema_source	''	Schema route: chewie-ns, external, or create.
--species	''	Species identifier for Chewie-NS and name used for a generated Prodigal training file.
--external_schema_dir	''	External schema to adapt with chewBBACA.
--ref_dir	''	Reference genome collection used for schema creation and/or model training.
--training_file	''	Prodigal .trn file or .fna genome used to produce one.
--model	maxst	Model type: maxst or imc.
--model_file	''	Existing .pkl model. Used when --new_model is false.
--new_model	false	Train a model from the reference collection.
--ref	''	Reference sample identifier used to order loci. Required for training.
--test	false	Run masking-based model evaluation.
--output	results	Top-level result directory.
--help	false	Print the command-line help message.

Citations

COMING SOON

Additionally, an extensive list of references for the tools used by the pipeline can be found in the CITATIONS.md file.

Contact

Did you find a bug? 🐛 Suggestion / Feedback / Feature request? 👨‍💻 Please visit GitHub Issues

For business inquiries or professional support requests, please feel free to contact us!