Robustness analysis of elementary flux modes generated by column generation

Area of Science:

• Systems Biology
• Metabolic Engineering
• Computational Biology

Background:

• Metabolic flux analysis (MFA) uses elementary flux modes (EFMs) to estimate metabolic fluxes.
• Data fitting in MFA relies on external flux measurements, which are prone to errors.
• Robust optimization can address these errors but often increases problem complexity.

Purpose of the Study:

• To develop a robust optimization framework for EFMs-based MFA that accounts for measurement errors.
• To demonstrate that the robust problem can be formulated as a convex quadratic programming (QP) problem.
• To apply column generation to the robust problem, avoiding explicit EFM enumeration and incorporating intervals for unmeasured metabolites.

Main Methods:

• Formulation of a robust optimization problem for EFMs-based MFA with bounded measurement errors.
• Application of a column-generation framework to solve the robust optimization problem efficiently.
• Inclusion of intervals for unmeasured metabolites within the column generation framework.

Main Results:

• The robust EFMs-based MFA problem with bounded measurement errors simplifies to a convex quadratic programming (QP) problem.
• Column generation effectively solves the robust problem without enumerating all EFMs.
• The case study indicated that non-robust solutions are near-optimal, suggesting measurement errors have limited impact.
• Incorporating intervals for unmeasured metabolites influenced the optimal solution.

Conclusions:

• Robust optimization for EFMs-based MFA is computationally tractable, solvable via convex QP and column generation.
• The developed method efficiently handles measurement errors and allows for the analysis of unmeasured metabolites.
• Findings suggest that typical measurement errors in MFA may not significantly alter the optimal flux distribution.