Section: New Results
Learning CRNs from data time series
Participants : François Fages, Jérémy Grignard [former&future Member] , Nicolas Levy, Julien Martinelli, Sylvain Soliman.
The problem of learning a mechanistic model from observed data is more difficult than learning a blackbox model fitting the data, due to the difference between causal relationships and correlations. In a biological context, learning a mechanistic model from experimental data would help understanding the underlying biological processes. To that end, considering multiple time series data generated by a hidden CRN from different initial states (by either stochastic or differential simulation), we develop a clustering-based algorithm for the inference of biological reaction networks. The output is a set of reactions which can be used to generate new traces. A model selection method is derived from these newly generated traces. We evaluate the performance of this algorithm on a range of models from Biomodels.