Model driven analysis and design of a lactate sensing genetic circuit

Optimization of a biosensor to differentiate between healthy and colon cancer environments

Project of Josia Pool for iGEM 2022 - supervised by Robert Smith.

Requirements

Python 3.6.9 to 3.10.6 (3.11 may work as well but will need updated packages.)

Setting up a virtual environment

I ended up using Venv, not anaconda for running python environments separate from the main installation. A virtual environment ensures you can use all dependencies and package versions that you need without creating conflicts with any other projects you might be working on. I would highly advise you to do the same.

Instructions on how to set up an environment can be found in "Using virtual environments.docx"

Running models and optimizing

Models are located in the models folder.

• ODEModelv2: Contains a general ODE model class with functions used in every other model that builds upon this. The code for the CRISPRi/asRNA model is also in here.
• Lactatemodel_succinate_p3: The model that was used to fit to the lactate biosensor data from the lab
• Bassystem_final: The combined model that was used to perform one-at-a-time sensitivity analysis
• Bassystem_GFP_pALPAGA: The combined model with the promoter for GFP switched from a constitutive one to a lactate dependent promoter, creating a feed forward loop motif.
• Bassystem_GFP_pALPAGA_sRNA: The combined model with a lactate dependent sRNA, decreasing leaky expression in low lactate concentrations.

In order to sample parameter sets used for fitting the model, check out this example:

#Add the models folder to path
import sys
sys.path.append("models/")

#Import the model
from Lactatemodel_succinate_p3 import Lactate_model_succinate_p3

#Initialize the model
model = Lactate_model_succinate_p3()

#Sample 1 million parameter samples
model.sample(1000000)

#Save samples
model.save_samples(1000000, "samplename") 
#The sample name can be anything to identify the samples by.
#Samples should be saved to a /samples folder (which should be created on initializing the model if its not available)
#The samples are split into 10 seperate files.

Simulation of the samples and optimization

One script is used for simulation of the samples: "optim_wrapper.py" This script is called using a command line terminal, an example is shown below:

python3 optim_wrapper.py --model p3model --method samplename --file 1

Arguments are as follows:

• --model: The model you would like to optimize, options are 'dCas9' for the CRISPRi model and 'p3model' for the lactate biosensor model.
• --method: The sample name which is the same when using save_samples.
• --file: The file number, from 1 to 10.

It will take a while to simulate all of these, so it is advised to leave the 10 sample lists running parallel on a server.

Once all samples are simulated, the sample lists are moved to samples/simulated and the files will now contain the scores of the simulated parameter sets. In order to process these simulated samples and continue with the minimization step, we use "process_optim_results.py". This script is also called using a command line terminal:

python3 process_optim_results.py --model p3model --method samplename --file 1000000 --cores 8

Arguments are as follows:

• --model: The model you would like to optimize, options are 'dCas9' for the CRISPRi model and 'p3model' for the lactate biosensor model.
• --method: The sample name which is the same when using save_samples.
• --file: The number of samples in total, as passed in save_samples.
• --cores: This script uses multiprocessing in order to speed up optimization, in order to take advantage of the ssb server. (This argument is optional, if this isnt used it will use 8 cores.)

This script will first report the best score and create some figures on the simulated latin hypercube samples, saved in lhs_results/. Afterwards the best 1000 parameter sets are optimized by L-BFGS-B, and the results of this are reported again. These results are saved in minimization_results/. My results can also be seen in these folders.

Description of jupyter notebook files

These are short descriptions for the jupyter notebook files in this repository. More details are given in the files themselves.

• BasDataExploration: This file processes all raw data from Bas and create plots to visualize the data.
• GrowthFit: This file contains the code for the preliminary fitting of the growth equation before fitting the entire lactate model
• Model_timedelay: This model contains the code used to evaluate if a time delay improves the fit of the CRISPRi/asRNA model
• BackgroundComparison: This model contains the code used to create Figure 18 and Table 1 of the thesis, comparing system improvements for the reduction of background expression.
• Test_alpaga: This code contains some plots which compare the Lactate model to the data. It also contains a plot of the 'desired' dose-response curve.

Final remarks

This documentation should give enough information to get started with my code, but if you have any questions you can always reach out to me