Reconstructing Kinetic Models for Dynamical Studies of Metabolism using Generative Adversarial Networks

Abstract

Kinetic models of metabolic networks relate metabolic fluxes, metabolite concentrations, and enzyme levels through well-defined mechanistic relations rendering them an essential tool for systems biology studies aiming to capture and understand the behavior of living organisms. However, due to the lack of information about the kinetic properties of enzymes and the uncertainties associated with available experimental data, traditional kinetic modeling approaches often yield only a few or no kinetic models with desirable dynamical properties making the computational analysis unreliable and computationally inefficient. We present REKINDLE (REconstruction of KINetic models using Deep LEarning), a deep-learning-based framework for efficiently generating large-scale kinetic models with dynamic properties matching the ones observed in living organisms. We showcase REKINDLE's efficiency and capabilities through three studies where we: (i) generate large populations of kinetic models that allow reliable in silico testing of hypotheses and systems biology designs, (ii) navigate the phenotypic space by leveraging the transfer learning capability of generative adversarial networks, demonstrating that the generators trained for one physiology can be fine-tuned for another physiology using a low amount of data, and (iii) expand upon existing datasets, making them amenable to thorough computational biology and data-science analyses. The results show that data-driven neural networks assimilate implicit kinetic knowledge and structure of metabolic networks and generate novel kinetic models with tailored properties and statistical diversity. We anticipate that our framework will advance our understanding of metabolism and accelerate future research in health, biotechnology, and systems and synthetic biology. REKINDLE is available as an open-access tool.

Rational strain design with minimal phenotype perturbation

“…These kinetic models consist of a system of nonlinear ordinary differential equations (ODEs) characterized by a set of kinetic parameters. To generate such models, we can use traditional kinetic modeling approaches such as MASSPy 17 , Ensemble Modeling 19 , ORACLE 15,18,28,29 , and machine learning empowered methods such as REKINDLE 21 and iSCHRUNK 20,22 .…”

“…To infer the values of missing kinetic parameters, researchers have traditionally employed parameter estimation [12][13][14] and Monte Carlo techniques [15][16][17][18][19] . Recently, there have been numerous efforts to use machine learning to accelerate the building of these models [20][21][22][23] .…”

Generative machine learning produces kinetic models that accurately characterize intracellular metabolic states

“…A parameterized kinetic model exhibits highly nonlinear but deterministic responses that depend on the intracellular state determined by the network topology and the integrated data. To capture this nonlinear behavior and determine kinetic parameters, we require function approximators with similar complexity, such as neural networks 32 . In RENAISSANCE, we iteratively optimize the weights of feed-forward neural networks (generators) using NES (Figure 1a) to obtain kinetic parameters leading to biologically relevant kinetic models, meaning that the metabolic responses obtained from these models have experimentally observed dynamics (Methods).…”

“…Recent efforts employing new tailor-made parametrization 27 and machine learning [32][33][34] improved the efficiency of constructing near-genome-scale kinetic models. Nevertheless, challenges remain regarding extensive computational time 27 and the need for training data from traditional kinetic modeling approaches [32][33][34] .…”

“…Recent efforts employing new tailor-made parametrization 27 and machine learning [32][33][34] improved the efficiency of constructing near-genome-scale kinetic models. Nevertheless, challenges remain regarding extensive computational time 27 and the need for training data from traditional kinetic modeling approaches [32][33][34] . Here, we present RENAISSANCE (REconstruction of dyNAmIc models through Stratified Sampling using Artificial Neural networks and Concepts of Evolution strategies), a machine learning framework that efficiently parameterizes biologically relevant kinetic models of metabolism without requiring training data.…”