A Novel Strategy for Dynamic Modelling of Genome-Scale Interaction Networks

Abstract

AbstractBackgroundModern medicine is equipped with huge amounts of big biological datasets and a wide range of computational methods to understand the molecular events underlying complex disorders. The recent availability of omics data allows a holistic view towards the interactions of various biomolecule types. However, the constructed maps are static, ignoring the dynamicity of molecular processes. On the other hand, the dynamic models of biological systems are commonly generated in small scales. Hence, the construction of large scale dynamic models that can quantitatively predict the time-course cellular behaviors is a big challenge. This study was aimed at developing a pipeline for automatic construction of such models from time-course experimental data.ResultsInformation of interactions between input genes is retrieved from SIGNORE 2.0 database and an interaction network is constructed which then is translated to biochemistry language and converted to a biochemical reactions network. In the next step, a large-scale ODE system is constructed by generating the ODE equivalent of each biochemical reaction. To estimate the kinetics parameters of the ODE model, a novel large-scale parameter approximation method is proposed. This method gives an estimation of system parameters by fitting model outputs to time-course experimental measurements. The total pipeline is provided as a MATLAB toolbox called SPADAN, standing for Systematic Protein Association Dynamic ANalyzer. Using multilayer time-series experimental data, the performance of the pipeline was checked by modeling 4379 regulatory interactions between 768 molecules in colon cancer cells exposed to chemotherapy agents.ConclusionStarting from time-series experimental data, SPADAN automatically constructs map of interactions, generates an ODE system, and performs a parameter approximation procedure. It constructs genome-scale dynamic models, filling the gap between large-scale static and small-scale dynamic modeling strategies. This simulation approach allows for holistic quantitative predictions which is critical for the simulation of therapeutic interventions in precision medicine.