coltonlloyd/optaux repository overview

⁠OptAux

Files and scripts needed to reproduce the figures in "Model-driven design and evolution of non-trivial synthetic syntrophic pairs"

⁠Docker

Project can be installed using docker by running from the docker folder

docker build -t optaux:everything .

this will install everything needed to reproduce most of the results outlined below including the community iJL1678b ME-model.

⁠Reproducing Figures/Tables

⁠OptAux simulations

Run python [optaux]/scripts_figures_and_tables/make_optaux_supplement.py

This runs OptAux algorithm in aerobic glucose minimal media conditions for all carbon containing exchange reactions in iJO1366. It will by default run OptAux for 4 competing_metabolite_uptake_thresholds (0, 0.01, 0.1, and 2) and output the results in supplement_1_optaux_solutions.xls as well as intermediate results as optaux_intermediate_trace_[threhold_value].xls

To process the results into an MSE summary spreadsheet and the set of EBC designs with the smallest number of knockouts, run output_optaux_summaries.xls

⁠Relative abundance approximations

Running python output_relative_abundance_results.py will:

1. Check if abundance_by_characteristic_[pair_name].csv and abundance_by_coverage_[pair_name].csv are located in [optaux]/scripts_figures_and_tables/relative_abundance/tables. If not, these CSVs will be created using the information in the read Bowtie2 alignment BAM files for all sequencing samples (not provided for now) and breseq generated mutation calls found in [optaux]/optaux/resources/resequencing_data/AUX [pair_name] Mutation Table.csv

   • Some values in the Mutation Tables for the characteristic strain mutations are missing because breseq rounds high/low mutation frequencies to 100% or 0%, respectively. These values are filled in with their frequencies in relative_abundance.py

2. Output bar charts of the characteristic mutation/alignment coverage base approximations of relative strain abundances in [optaux]/scripts_figures_and_tables/relative_abundance/figures as well as a comparison of the predictions based on each method.

⁠Duplications

Read coverages are plotted using the alignment files produced from breseq (Not provided due to filesize limits on github. They can be obtained by contacting [email protected]⁠ for access to the alignment files or accessing the raw reads hosted on the Sequence Read Archive under accession no. SRP161177⁠).

Running python output_duplications.py will:

1. Check if [optaux]/scripts_figures_and_tables/duplications/[pair_name]_coverage_dict.json exists. If not it will produce this dictionary using the bam files output from breseq.

2. Find the genes with >80% of their base pairs above the 1.25x fit mean cutoff. These are compiled and output in [optaux]/scripts_figures_and_tables/duplications/duplicated_genes

3. Output the plots used to create Figure 7 and the supplementary figures in [optaux]/scripts_figures_and_tables/duplications/

⁠Community Modeling Sims and Plotting

The iJL1678b ME-model (constructed using COBRAme/ECOLIme v0.0.9) is provided as json files with two different k_eff parameter sets:

• iJL1678b.json: The model with default parameters

• iJL1678b_null_keffs.json: The model with k_effs values set to those obtained from Dividi et. al.⁠. Metabolic k_effs not included in this dataset were imputed with the median value, 6.2 s^-1. Non-metabolic reactions outside the scope of this dataset were set to 65 s^-1 as in the default model.

The key results from the study can be reproduced with the following

1. Build iJL1678b-community models with: python [optuax]/optaux/me_community/make_me_communityl.py.

   • This will make community models based on the two ME-models described as well as community model with all k_effs set to 65 s^-1.

2. All of the simulations needed to reproduce Figures 8 and 9 can then be ran with: python run_community_me_sims.py

   • This code uses python's multiprocessesing functionality with 2 processes by default. To speed up these simulations, more processes can be used. A single simulation typically requires 4-8 gb of RAM to solve with qMINOS.
   • This will output the results into [optaux]/scripts_figures_and_tables/community_sims_output_[null/65/default]_keffs. Alternatively, tar.gz files are containing the output of these simulations are already included in this package. Note: If using these files instead of running new simulations you must run python unpack_community_me_sims.py to unpack the tar.gz files in order to plot.

To recreate Figure 9, part of Figure 8, and the supplementary community ME-figures:

1. To plot community growth rates for varying strain abundances run: python [optaux]/scripts_figures_and_tables/output_computed_community_growth_rates.py

2. To plot metabolite cross-feeding for varying strain abundances run: python [optaux]/scripts_figures_and_tables/output_computed_metabolite_crossfeeding.py

3. To plot community growth rates for substrate limited ME-model and M-model run: python [optaux]/scripts_figures_and_tables/output_glucose_limited_me_m_comparison.py

4. To output the supplement community growth comparison between steadycom and jointfba M-model: python [optaux]/scripts_figures_and_tables/output_steadycom_jointfba_comparison.py

⁠Software

The following software and versions were used for publication:

• Python 3.6
• An MILP solver. The scripts here use Gurobi by default.
• COBRApy v0.5.11
• COBRAme⁠ v0.0.9
• ECOLIme⁠ v0.0.9
• solvemepy⁠ v1.0.1
  • Including the qMINOS solver
• pysam v0.14.1
• openpyxl v2.3.2
• pandas v0.22.0
• matplotlib v2.0.2
• numpy v1.14.2
• scipy v0.19.0
• Biopython v1.66
• Seaborn v0.7.1