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.
展开▼