Machine learning detection of unstable antibiotic heteroresistance in E. coli

This repository contains code and certain types data for the project published in %journalname%.

The pipelines were created using Snakemake v7.32.4

Data analysis was performed using R v4.4.1 and machine learning was performed using tidymodels v1.2.0.

How to run the main analysis

First, ensure that the raw (short and long) reads are placed in resources/data_raw/{strain}/short/ and resources/data_raw/{strain}/long/ directories inside the project's directory.

After this is done, you can run the main pipeline (genome assembly, anotation of resistance genes, repeats and insertion sequences):

a) with this command to use conda environments:

# navigate to the project's directory
snakemake --snakefile workflow/snakefile.smk --configfile workflow/config.yaml --use-conda 

or

b) with this command to use Apptainer containers instead:

# navigate to the project's directory
snakemake --snakefile workflow/snakefile.smk --configfile workflow/config.yaml --use-singularity

for more details on installation of snakemake and available options (number of threads, running on computer clusters etc), see the official Snakemake documentation

Then you can run the three R notebooks:

1. generate features table: notebooks/modelling/features.qmd (it is already available in notebooks/modelling/data/features_strain.csv)
2. (optional) exploratory data anlysis: notebooks/modelling/EDA.qmd
3. run training and validation:notebooks/modelling/training_and_validation.Rmd
4. comparison and analysis of the models: notebooks/modelling/models_analysis.Rmd

To install the same versions of R packages as were used in these notebooks, install renv package first and then run renv::restore() (here you can find renv documentation).

How to run the additional analyses

Phylogenetic analysis

snakemake --snakefile workflow/phylogeny.smk --configfile workflow/config_phylogeny.yaml --use-conda 

NB: 31 reference strain are required (see the publication for reference numbers).

Analysis of the HR mutants

# analysis of the HR mutants
snakemake --snakefile workflow/mutants.smk --configfile workflow/config_mutants.yaml --use-conda 

Raw data availability

The raw sequencing reads used in this project are available from NCBI's SRA under BioProjects PRJNA1165464 and PRJNA1083935.

Models and features table

The pre-compiled features table is available here: notebooks/modelling/data/features_strain.csv

The final models (trained LLR and GBT) are available here: notebooks/modelling/models/llr_final.rds & notebooks/modelling/models/gbt_final.rds

Rule graphs

1. The main analysis
2. HR mutants analysis