Endy:Chassis engineering/Computational load modeling

Introduction

Based on my thesis committee meeting and subsequent conversations with Drew, we decided that aspects of my project would benefit from some modeling work. Firstly, we believe that modeling could better inform the design of feedback control for the dedicated systems of VM2.0. Secondly, we currently lack a clear model of how the demands of an engineered system vary with the system parameters, such as DNA copy number, promoter PoPS, RBS RiPS etc.

Initial questions

1. What goals should I set for the modeling work to get some benefit from it without devoting a long period of time to it?
2. What modeling approach should I adopt?

These questions are discussed below:

Modeling Objectives

• Build a simple model of gene expression that considers the finite resources of the cellular chassis and the fraction of those resources consumed by the gene expression process.
• Construct the simple model of gene expresssion in a modular fashion such that it can be used to model a genetic network.
• Use the model to test the network dynamics of a number of possible feedback control configurations for VM2.0.

While the third of these goals might be the most immediately important it might be worth proceeding in the order listed here to have a more powerful tool in the longer term. The objectives I choose will partly determine the best modeling approach to use.

Modeling Approach

Two main modeling approaches exist - deterministic, continuous models based on differential equations and discrete, stochastic simulation of individual biochemical reactions. Both of these approaches offer some advantages and disadvantages for me.

Continuous deterministic modeling

• I have more experience with this.
• Analytical solutions give greater insight.
• Can do this in Matlab, quite quickly.
• Computationally efficient - easier to get to steady state of the system which is probably the most interesting state.

• Hard to model some of the things I'd like to model. For example if I want to explicitly model the numbers of ribosomes on the transcript I need to create species and reactions like the following mRNA.(Ribosome)_n+Ribosome -> mRNA.(Ribosome)_n+1
• The drawbacks of continuous modeling when the molecule numbers are potentially small.
  • A less expicit, but probably adequate approach is just to include the following species - RBS, RBS.Ribosome, Ribosome_elong. This doesn't explicitly model an mRNA with varying numbers of ribosomes on it but the number of Ribosome_elong would implicitly tell me how many ribosomes are on each transcript. Thanks Sri.

Discrete stochastic simulation

• TABASCO
• Explicitly models a gene expression system's usage of the chassis' resources.
• Accurate simulation of dynamics of small molecule numbers.

• I think this would require me to simulate the time evolution of the system even if I only wanted to look at the steady state. I would also have to run many simulations.
• java.language.ignorance.Barry
• I might gain less intuition and insight from this approach since there would be no analytical solutions. I could try and do both; simulate discretely while at the same time analytically solving some parts of the system.

If the dynamics of VM2.0 are the most important thing to get out of the modeling, then a continuous model is probably the best way to go, if I want to do a thorough job of the modeling that lets me look closely at the demand of whatever system I'm modeling, then a discrete simulation might be better.