HetaProject – a Single Open Source Platform for Modeling, Simulation and Parameters Estimation in QSP
Ivan Borisov, Evgeny Metelkin
InSysBio, Moscow, Russia
Objectives: Pharmacological modeling is evolving towards large-scale dynamical models [1]. A typical QSP project may include thousands of molecular components and several drugs. Such project is usually carried out by a group of researchers with different fields of expertise. Models’ complexity increases the requirements for computational procedures including solving large-scale differential equations systems, Monte-Carlo simulations, parameter estimation, uncertainty quantification etc. Software, which could establish the linkage between emerging QSP frameworks and fast computational methods, is of utmost importance for the field.
Methods: Heta language is a Domain Specific Language (DSL) for development and annotation of dynamical models in QSP projects. HetaCompiler [2] allows the users to convert Heta-based models to one of the formats: DBSolve, SimBiology, mrgsolve, SBML, Julia. Models converted to Julia can be simulated and fitted to experimental data with HetaSimulator – Julia-based simulation and parameters estimation platform. Julia allows code precompilation and simple framework for parallel simulations. The package utilizes features of SciML ecosystem, like access to 300+ ODE solvers, automatic differentiation, symbolic Jacobian derivation, etc. HetaSimulator can be integrated with the LikelihoodProfiler package [3] to perform identifiability and predictability analysis.
Results: HetaProject includes a DSL, HetaCompiler and HetaSimulator, which form a full-cycle modeling and simulation environment. It is intensively used in both research and commercial QSP projects at InSysBio: Alzheimer-consortium platform, Immune Response Template, PK/RO simulator for anti-PD-1 mAbs. Heta DSL and Heta Compiler allow the user to write complex QSP models and convert them to popular modeling formats, while HetaSimulator proposes an interface to define a simulation or a parameter estimation problem, which supports multi-conditional regimes, different experimental data inputs and distribution types. Both modeling and simulation operations can be performed in a single interface of VSCode.
Conclusions: HetaProject is a single platform for models development and validation in the QSP field. It was released as an ecosystem of open-source software packages (https://github.com/hetalang/) and it can be can be used as the modeling and simulation environment in QSP project of any size and complexity. A number of use-case models illustrate the stages of models development and validation, which can be performed in a single framework.
References:
[1] Knight-Schrijver et al. Computational and Structural Biotechnology Journal (2016) 14, 363-37
[2] E. Metelkin. "Heta compiler" is a framework for the development and management of Quantitative Systems Pharmacology modeling platforms. American Conference on Pharmacometrics 11, November 2020 http://dx.doi.org/10.13140/RG.2.2.18555.13602
[3] I. Borisov, E. Metelkin “Confidence intervals by constrained optimization — An algorithm and software package for practical identifiability analysis in systems biology” PLoS Computational Biology 16(12):e1008495 http://dx.doi.org/10.1371/journal.pcbi.1008495