BayesianOccam’s Razorto Optimize Models for Complex Systems

Abstract

Developing and optimizing models for complex systems poses challenges due to the inherent complexity introduced by multiple types of input information and sources of uncertainty. In this study, we utilize Bayesian formalism to analytically examine the propagation of probability in the modeling process and propose quantitative assessments for it. Upon which, we develop a method for optimizing models for complex systems by (i) minimizing model uncertainty; (ii) maximizing model consistency; and (iii) minimizing model complexity, following the BayesianOccam’s razorrationale. We showcase the benefits of this method by optimizing the modeling of the dynamic system of glucose-stimulated insulin secretion in pancreaticβ-cells, leading to an optimized model that demonstrates better alignment with experimental observations compared to the non-optimized one. We anticipate that this method will facilitate the construction of accurate, precise, and sufficiently simple models for diverse complex systems. It is implemented in our open-source softwareIntegrative Modeling Platform(IMP), ensuring its broad applicability.