Assessing and Resolving Model Misspecifications in Metabolic Flux Analysis

Area of Science:

• Metabolic Engineering
• Systems Biology
• Biotechnology

Background:

• Metabolic flux analysis (MFA) is crucial for metabolic engineering, often using overdetermined stoichiometric models.
• Model errors, especially in the stoichiometric matrix, are a significant but understudied problem in MFA.
• Accurate intracellular flux distribution is key to understanding cellular metabolism.

Purpose of the Study:

• To evaluate statistical tests for detecting model misspecifications in overdetermined MFA.
• To propose and validate a method for correcting missing reactions in MFA models.
• To enhance the reliability of flux estimates in metabolic engineering.

Main Methods:

• Applied statistical tests (Ramsey's RESET, F-test, Lagrange multiplier) to linear least square regressions in MFA.
• Utilized an iterative F-test procedure to identify and correct for missing reactions.
• Tested the methods on Chinese hamster ovary and random metabolic networks.

Main Results:

• Statistical significance in regression does not ensure accurate flux estimates.
• Removing low-flux reactions can introduce substantial bias in flux estimations.
• The F-test effectively detected missing reactions, and the iterative procedure robustly resolved their omission.

Conclusions:

• Statistical tests are valuable tools for assessing and correcting model misspecifications in overdetermined MFA.
• The proposed iterative method improves the robustness of metabolic models.
• This work provides a systematic approach to enhance the accuracy of metabolic flux analysis.