Architectural Design of Chemical and Biological Pathways through Parameter Sensitivity Oriented Mixed Integer Formulations

Abstract

A mixed integer nonlinear programming (MINLP) formulation is developed for the automatized design of chemical and biological pathways. The proposed formulation is solved using rigorous algorithms and enables the selection of reactions whose rate constants have the highest sensitivity. Unlike the traditional parameter estimation approach, the training performance is introduced as constraint which is pre-defined to balance the architectural simplicity and fitting performance. Adjoint sensitivity expressions are additionally solved simultaneously in the formulation and the resulting parameter sensitivities are included in the objective. Binary variables are introduced in the linking constraints for the selection of a reaction subset from a more complex pathway, also tightening the search space for the optimization algorithm. The approach is implemented on a complex chemical/biological pathway which processes glucose as a primary feedstock to produce hydroxymethylfurfural (HMF). The reduced pathway prediction bounds are narrower ensuring a similar mean predictions, despite 25% reduction in the superstructure. © 2023 Elsevier B.V., All rights reserved.