Posters
« Back
Parameters Estimation, Identifiability and Predictability Analysis in Systems Biology with HetaSimulator and LikelihoodProfiler
EP39335
Poster Title: Parameters Estimation, Identifiability and Predictability Analysis in Systems Biology with HetaSimulator and LikelihoodProfiler
Submitted on 29 Sep 2022
Author(s): Ivan Borisov, Evgeny Metelkin
Affiliations: InSysBio
This poster was presented at PAGE 2022
Poster Views: 247
View poster »


Poster Information
Abstract: Introduction: Systems Biology and QSP Modeling workflow consists of time consuming and computationally demanding procedures, which include developing a model and improving its reliability and predictability. Currently models development, parameters estimation and identifiability analysis are often performed with different software tools, which results in compatibility issues and project delays. HetaSimulator [1] and LikelihoodProfiler [2] Julia-based open-source software packages constitute a software environment, which can be used both for editing a model, estimating the unknown parameters and performing indentifiability and predictability analysis in the single interface of VSCode.

Objectives: The goal of the study is to examine the application of both HetaSimulator and LikelihoodProfiler packages to simulation, parameters estimation, identifiability and predictability analysis of the model describing intracellular dynamics of influenza virus [3].

Methods: Heta DSL together with HetaSimulator and LikelihoodProfiler allows the users to develop a kinetic model in a human-readable format. It provides integration between models development and simulations and gives the user an access to the latest methods and tools for solving large-scale ODE systems, fitting the model to multi-conditional experimental datasets and performing both identifiability and predictability analysis. Original influenza virus model and experimental conditions were reproduced from the publication [3] and converted to Heta format [1]. Parameters estimation was performed with HetaSimulator on the basis of maximum likelihood estimation (MLE) approach. CICO algorithm based on Profile Likelihood method and proposed in LikelihoodProfiler package [2] was used to reveal identifiable parameters and construct Prediction and Validation Confidence Bands. Both packages rely on Julia SciML ecosystem, which provides access to over 300 ODE solvers and optimization algorithms [5].

Results: Models parameters were estimated based on the available experimental data with HetaSimulator and correspond with the results published in the original paper [3]. LikelihoodProfiler algorithm [2] initially designed to test parameters identifiability was proved to be efficient for more general use case of predictability analysis. Both identifiable and non-identifiable parameters of the model were obtained and predictability analysis revealed how identifiability affects the uncertainty in models predictions. Those uncertainties were estimated with Prediction and Validation Confidence Bands. All steps of models development, analysis and validation were performed in VSCode environment.

Conclusions: HetaSimulator and LikelohoodProfiler software packages allow the users to efficiently develop, simulate and analyze Systems Biology and QSP models in a single framework of VSCode. This open source ecosystem of packages can be can be used as a modeling and parameters estimation and uncertainty analysis environment in Systems Biology and QSP.
Summary: HetaSimulator and LikelohoodProfiler software packages allow the users to efficiently develop, simulate and analyze Systems Biology and QSP models in a single framework of VSCode. This open source ecosystem of packages can be can be used as a modeling and parameters estimation and uncertainty analysis environment in Systems Biology and QSP.References: [1] HetaProject: https://hetalang.github.io/
[2] Borisov I, Metelkin E. (2020) Confidence intervals by constrained optimization-An algorithm and software package for practical identifiability analysis in systems biology. PLOS Computational Biology 16(12):e1008495. doi: 10.1371/journal.pcbi.1008495.
[3] Heldt FS, Frensing T, Pflugmacher A, Gröpler R, Peschel B, et al. (2013) Multiscale Modeling of Influenza A Virus Infection Supports the Development of Direct-Acting Antivirals. PLOS Computational Biology 9(11): e1003372. https://doi.org/10.1371/journal.pcbi.1003372
[4] Metelkin, E., (2021). Heta compiler: a software tool for the development of large-scale QSP models and compilation into simulation formats. Journal of Open Source Software, 6(67), 3708, https://doi.org/10.21105/joss.03708
[5] Rackauckas, C. and Nie, Q., (2017). DifferentialEquations.jl – A Performant and Feature-Rich Ecosystem for Solving Differential Equations in Julia. Journal of Open Research Software,
Report abuse »
Questions
Ask the author a question about this poster.
Ask a Question »

Creative Commons