ABC(SMC)2: Simultaneous Inference and Model Checking of Chemical Reaction Networks

Abstract

We present an approach that simultaneously infers model parameters while statistically verifying properties of interest to chemical reaction networks, which we observe through data and we model as parametrised continuous-time Markov Chains. The new approach simultaneously integrates learning models from data, done by likelihood-free Bayesian inference, specifically Approximate Bayesian Computation, with formal verification over models, done by statistically model checking properties expressed as logical specifications (in CSL). The approach generates a probability (or credibility calculation) on whether a given chemical reaction network satisfies a property of interest.