Genomics-accelerated discovery of diverse fungicidal bacteria

This repository holds data and code necessary to reproduce the results in the manuscript "Genomics-accelerated discovery of diverse fungicidal bacteria".

The "raw" data includes:

• Information about each isolate, including taxonomic assignments and screening results isolates.tsv
• BGCs identified in each isolate, and the results of clustering them into families bgc_clustering_results.tsv
• List of BGCs used to query the collection for the in vivo experiment prioritization_exp_features.tsv
• List of BGCs from the AgBiome collection that clustered with the query BGCs collection_query_results.tsv
• Taxonomic assignments for isolates in the AgBiome collection that contained hits to the query query_taxonomy.tsv

To reproduce figures and results, run the following command:

make all

Here's a full list of make commands:

 make...
 help                                        : Show this help message
 build-image                                 : Build the docker image for notebooks
 build-graphlan                              : Build the docker image for graphlan
 build-alluvial                              : Build the docker image for ggalluvial
 run-notebook                                : Run the notebook server
 stop-notebook                               : Stop the notebook server
 get-url                                     : Get the URL of the notebook server
 data/category_covariance.png                : Produce the metadata covariance figure
 data/table2.tsv                             : Produce Table 2 and related feature selection analyses
 images/model_cv.png                         : Perform cross-validation analysis to compare modeling approaches
 data/sa_rf_model_pickle                     : Train the machine learning models
 data/collection_query_pa.tsv                : Evaluate performance of ML models on validation set
 generate-cladograms                         : Produce the cladogram from the ms
 generate-cladograms                         : Produce the cladogram from the ms
 clean                                       : delete all images and derivations, keeping only original data files
 all                                         : reproduce all figures and analyses

	Figure	Legend
Full Cladogram
Alluvial of all screened isolates
Only Actives
Metadata Co-Variance
BGC Co-Occurrence
ML Crossvalidation
Effect of Oversampling on RF
ML Performance and Taxonomic Signal
Excluding taxa from training set
Model performance in validation experiment
Model performance in compared to random draws

In brief, the make all command runs the following notebooks:

• summaries_and_table1.ipynb generates Table 1 and Supplemental Figure 2
• table2_bgc_feature_selection.ipynb generates Table 2, Supplemental Table 1, Supplemental Figures 3 and 4.
• model_cross_validation.ipynb performs the ML cross-validation experiment, Figure 3 and Supplemental Figure 6
• prioritization_exp_model_training.ipynb generates Supplemental Figure 5, and trains the models used in the in vivo experiment
• model_performance.ipynb evaluates the results of the in vivo experiment, generates Figure 4.
• figure2_cladogram.ipynb generates the GraphLan files to produce the cladograms in Figure 2 and Supplemental Figure 1.
• all_screened_alluvial.R generates the alluvial plot of all screened isolates in Figure 2.

The code and data in this notebook fit together like this: