pySODM: Simulating and Optimizing Dynamical Models in Python 3

TL;DR

pySODM is a versatile Python framework that simplifies constructing, simulating, and calibrating complex dynamical models across scientific disciplines, demonstrated through biochemical and epidemiological case studies.

Contribution

The paper introduces pySODM, a general Python toolkit that streamlines the development, simulation, and calibration of dynamical systems with n-dimensional states.

Findings

Successfully modeled enzymatic reaction yields for process design.

Accurately forecasted influenza epidemic evolution with limited data.

Demonstrated cross-disciplinary applicability of the framework.

Abstract

In this work, we present our generic framework to construct, simulate, and calibrate dynamical systems in Python 3. Its goal is to reduce the time it takes to implement a dynamical system with $n$-dimensional states represented by coupled ordinary differential equations (ODEs), simulate the system deterministically or stochastically, and, calibrate the system using n-dimensional data. We demonstrate our code's capabilities by building three models in the context of two case studies. First, we forecast the yields of the enzymatic esterification reaction of D-glucose and lauric acid, performed in a continuous-flow, packed-bed reactor. The model yields a satisfactory description of the reaction yields under different flow rates and can be applied to design a viable process. Second, we build a stochastic, age-stratified model to make forecasts on the evolution of influenza in Belgium during the 2017-2018 season. Using only limited data, our simple model was able to make a fairly accurate assessment of the future course of the epidemic. By presenting real-world case studies from two scientific disciplines, we demonstrate our code's applicability across domains.