Partial inhibition and bilevel optimization in flux balance analysis

Area of Science:

• Computational Biology
• Systems Biology
• Metabolic Engineering

Background:

• Flux Balance Analysis (FBA) often uses simplified ON/OFF models for network perturbations.
• Existing models struggle to realistically represent partial drug inhibition in complex biological systems.
• Bilevel optimization is crucial for tasks like identifying optimal drug combinations or network perturbations.

Purpose of the Study:

• To develop a bilevel optimization formulation that accurately models partial drug inhibition.
• To overcome the limitations of oversimplified ON/OFF descriptions in metabolic network analysis.
• To enable more realistic and comprehensive investigations of drug effects on cellular metabolism.

Main Methods:

• Introduced a novel bilevel optimization formulation allowing for continuous modulation of drug effects.
• Modeled drug inhibition using a convex combination of Boolean variables, preserving linearity.
• Applied the method to the core metabolism of E. coli for a case study on multi-drug treatment optimization.

Main Results:

• Successfully overcame the limitations of ON/OFF modeling in bilevel optimization problems.
• Demonstrated the ability to find a wider range of drug combinations compared to traditional methods.
• The approach identified synergistic drug combinations for modulating specific metabolic reactions with minimal network perturbation.

Conclusions:

• The presented method effectively handles partial inhibition in bilevel optimization without sacrificing linearity.
• Achieved reasonable computational performance on large-scale metabolic networks.
• This fine-grained perturbation modeling expands therapeutic options for selective cellular metabolism manipulation.