Diagnosing Productivity Drops in Cell Factories with Mixed Integer Linear Programming

April 12, 2023
At the Merantix Research Day 2023, Pierre Salvy, Head of Engineering at Cambrium, discussed how his team has been using mixed integer linear programming to diagnose productivity drops in cell factories.

The Merantix Research Day is a yearly event where all of our Merantix ventures come together to share what they are working on and highlight technical achievements that have been made. This article series highlights a few of the talks given at the 2023 edition of our Research Day, and was written with the help of Open AI’s Whisper and ChatGPT.

The Need for Models in Biology

At the Merantix Research Day 2023, Pierre Salvy, Head of Engineering at Cambrium, discussed how his team has been using mixed integer linear programming to diagnose productivity drops in cell factories. Salvy focused on three main aspects: the relationship between optimization and biology, formulating a model, and using the model to explain lab observations. Studying complex biological processes using computers is essential due to the impossibility of directly observing what happens inside a cell.
By developing models, researchers can gain insights on the inner state of the cells, which helps guide experimentation and prevent blind manipulation. Metabolism can be represented as a set of constraints, with linear differential equations describing the variation of chemical concentrations. The models, constructed from these constraints, play a crucial role in understanding cell behaviour in different scenarios and can provide information on the hot spots of potential issues.

Formulating a Model with Protein Constraints and General Structure

Adding protein constraints to the model allows for the inclusion of enzymes, specialised proteins that facilitate reactions. By including these constraints, researchers can better understand the relationship between enzyme concentrations and reaction rates. In particular, the model can describe how the rate of a reaction is lower than a constant times the concentration of the enzyme that drives this reaction. The general structure of the model consists of metabolism on one side and gene expression on the other. The metabolism side represents the conversion of chemicals such as A to B and B to C, ultimately leading to amino acids. These amino acids then make enzymes and proteins that regulate the production of chemicals A, B, C, and D, all of which contribute to cell growth. The gene expression side is more complex and nonlinear, with integer variables that can make optimization difficult.

Optimization Techniques and Model Construction

Solving the complex optimization problem presented by the model requires a combination of linear and convex relaxations, as well as ‘semismatch’ heuristics. By applying these techniques, the team constructed a model representing the cell at the metabolism and gene expression levels, containing thousands of constraints and variables. The optimization techniques allow the otherwise complicated problem to be solved in a matter of minutes rather than hours, making it more practical for real-world applications.

Investigating the Impact of Food Sources and Oxygen Consumption

Using the model, the researchers investigated the impact of different feedstock on cell growth. The model demonstrated the non-linear relationship between food intake and protein production capacity, as well as how different food sources led to distinct metabolic behaviors within the cell. Through the model, the researchers were able to visualize the metabolism maps and understand the impact of various food sources on cellular energy production and detoxification processes.

The Power of Collaboration and Optimization in Bioproduction

The collaboration between biology and computational research is essential for understanding and engineering cells to perform better. Mixed integer linear programming offers a powerful tool to diagnose and address productivity drops in cell factories. As technology advances, these models will become even more precise, allowing for further optimization and improvements in bioproduction processes. The insights gained from these models not only help in identifying potential problems but also pave the way for innovative solutions in the field of biotechnology, ultimately contributing to increased productivity and efficiency in bioproduction.

Read more