Chemical space visualisations for bacterial growth screen compounds

This repository contains the scripts and output for the visualisations of the chemical space coverage for the compounds screened for their impact on bacterial growth. The results are presented in Lindell, Kamrad & Roux et al.

Set up environment

To create a suitable environment to run the script, we suggest using conda/mamba. The following commands will create a working environment with the dependencies installed.

conda create -n compound_visualisation python=3.10
conda activate compound_visualisation
mamba install -y -c conda-forge -c bioconda umap-learn pandas numpy scipy \
    matplotlib seaborn chemplot bokeh==2.4.3 \
    r-ggplot2 r-optparse r-dplyr r-devtools
R -e "devtools::install_github('VPetukhov/ggrastr')"

To generate the UMAP plot, run:

python scripts/chemical_space_UMAP.py -pc raw/CID-SMILES_random_25K.tsv -sc raw/compound2smile.tsv --log umap.log --output-prefix results/umap

To plot the Tanimoto distance between PubChem compounds and screened compounds, run the following. Note that results/umap_min_jaccard.tsv and results/umap_min_pollutant_drug_jaccard.tsv are files created by scripts/chemical_space_UMAP.py.

scripts/plot_tanimoto.R --min-jacc-inf results/umap_min_jaccard.tsv --min-jacc-pol-drug-inf results/umap_min_pollutant_drug_jaccard.tsv --plot-prefix results/plot