In this paper, a computational framework is proposed that merges mechanistic modeling with deep neural networks obeying the Systems Biology Markup Language (SBML) standard. Over the last 20 years, the systems biology community has developed a large number of mechanistic models that are currently stored in public databases in SBML. With the proposed framework, existing SBML models may be redesigned into hybrid systems through the incorporation of deep neural networks into the model core, using a freely available python tool. The so-formed hybrid mechanistic/neural network models are trained with a deep learning algorithm based on the adaptive moment estimation method (ADAM), stochastic regularization and semidirect sensitivity equations. The trained hybrid models are encoded in SBML and uploaded in model databases, where they may be further analyzed as regular SBML models. This approach is illustrated with three well-known case studies: the Escherichia coli threonine synthesis model, the P58IPK signal transduction model, and the Yeast glycolytic oscillations model. The proposed framework is expected to greatly facilitate the widespread use of hybrid modeling techniques for systems biology applications.

Hybrid modeling workflows combining machine learning with mechanistic process descriptions are becoming essential tools for bioprocess digitalization. In this study, a hybrid deep modeling method with state–space reduction was developed and showcased with a P. pastoris GS115 Mut+ strain expressing a single-chain antibody fragment (scFv). Deep feedforward neural networks (FFNN) with varying depths were connected in series with bioreactor macroscopic material balance equations. The hybrid model structure was trained with a deep learning technique based on the adaptive moment estimation method (ADAM), semidirect sensitivity equations and stochastic regularization. A state–space reduction method was investigated based on a principal component analysis (PCA) of the cumulative reacted amount. Data of nine fed-batch P. pastoris 50 L cultivations served to validate the method. Hybrid deep models were developed describing process dynamics as a function of critical process parameters (CPPs). The state–space reduction method succeeded to decrease the hybrid model complexity by 60% and to improve the predictive power by 18.5% in relation to the nonreduced version. An exploratory design space analysis showed that the optimization of the feed of methanol and of inorganic elements has the potential to increase the scFv endpoint titer by 30% and 80%, respectively, in relation to the reference condition.

In this paper we propose a computational framework that merges mechanistic modeling with deep neural networks obeying the Systems Biology Markup Language (SBML) standard. Over the last 20 years, the systems biology community has developed a large number of mechanistic models in SBML that are currently stored in public databases. With the proposed framework, existing SBML mechanistic models may be upgraded to hybrid systems through the incorporation of deep neural networks into the model core, using a freely available python tool. The so-formed hybrid mechanistic/neural network models are trained with a deep learning algorithm based on the adaptive moment estimation method (ADAM), stochastic regularization and semidirect sensitivity equations. The trained hybrid models are encoded in SBML and stored back in model databases, where they can be further analyzed as regular SBML models. The application of this approach is illustrated with three well-known case studies: the threonine synthesis model in Escherichia coli, the P58IPK signal transduction model, and the Yeast glycolytic oscillations model. The proposed framework is expected to greatly facilitate the widespread use of hybrid modeling techniques for systems biology applications.