AutoCD: Automated synthetic microbial Community Designer

Automated equation generation of candidate models for synthetic microbial communities and performs model selection using ABC SMC for customisable objective population behaviour.

AutoCD builds a equations for candidate model systems that can simulate the growth of bacteria capturing important interactions such as competition for nutrients, communication through quorum sensing and amensalism via bacteriocins. AutoCD performs model selection for a desired population level behaviour. AutoCD is intended to aide in the design of synthetic microbial systems, identfying promising candidate network topologies and essential parameters for controlling a microbial community.

AutoCD is written in C++ and Python 3.4+, data analysis scripts require R.

This software has been tested on Linux and designed for HPC, using OpenMP for parallel simulations.

Requirements

Python dependencies:

matplotlib, sympy, seaborn, pandas, scipy, numpy, PyYAML. See requirements.txt for details of versions used and tested with.

C++ dependencies:

g++ (tested with g++-8) Boost >= v1.67 OpenMP

Boost must be installed for the version of python being used to run AutoCD. Ensure boost has been correctly added to paths. Run build_test.sh to test on a simple build example and identify issues with library flags and ensuring the correct Python includes. Changes should be made to build_pop_modules.sh which is used to the build necessary Boost Python libarary for a newly generated model space.

R packages:

ggplot2, ggisoband, gridExtra, sm

Usage

Experiments are run using run_AutoCD.py. The Model Space Generator, Model Selection and Data Analysis parts of AutoCD are called using optional config file arguments

Model Space Generator

The --MSG_config is an optional command-line argument is used to invoke the Model Space Generator and requires a config file as input.

Examples running Model Space Generator for two and three species model spaces:
python3 run_AutoCD.py --MSG_config ./config_files/MSG_two_strain.yaml
python3 run_AutoCD.py --MSG_config ./config_files/MSG_three_strain.yaml

MSG_two_strain.yaml and MSG_three_strain.yaml contain examples of how to specify a model space config file.

ABC Model Selection

The --ABC_config is an optional command-line argument used to invoke the model selection using ABC SMC or ABC rejection algorithms, requiring a config file as input.

ABC_two_strain.yaml and ABC_three_strain.yaml contain examples of how to specify an ABC config file. --exp_num is used to differentiate betweet repeats of using the same input files. Each experiment is stored in a subdirectory, all subdirectories are used by data analysis scripts

Examples running ABC SMC model selection for two and three species model spaces:
python3 run_AutoCD.py --ABC_config ./config_files/ABC_two_strain.yaml --exp_num 0
python3 run_AutoCD.py --ABC_config ./config_files/ABC_three_strain.yaml --exp_num 0

Data Analysis

The --DA_config is an optional command-line argument used to invoke data analysis of a set of finished ABC experiments, requiring a config file as input.

Examples running data analysis for two and three species model spaces:
python3 run_AutoCD.py --DA_config ./config_files/DA_two_strain.yaml --exp_num 0

Example

Here is a typical bash script that first generates a model space, runs four repeats of ABC SMC model selection, then performs data analysis on these repeat experiments

#!/bin/bash

# Set number of repeat experiments
n_repeats=4

# Run model space generator
python3 run_AutoCD.py --MSG_config ./config_files/MSG_two_strain.yaml

# Conduct n repeat model selection experiements
for i in $(seq 0 $n_repeats); do 
	python3 run_AutoCD.py --ABC_config ./config_files/ABC_two_strain.yaml --exp_num $i
	echo $i; 
done

# Run data analysis
python3 run_AutoCD.py --DA_config ./config_files/DA_two_species.yaml