Quantifying structural uncertainty in chemical reaction network inference

Abstract: Dynamical systems in chemistry and biology are complex, and one often does not have comprehensive knowledge about the interactions involved. Chemical reaction network (CRN) inference aims to identify, from observing species concentrations over time, the unknown reactions between the species. Existing approaches largely focus on identifying a single, most likely CRN, without addressing uncertainty about the resulting network structure. However, it is important to quantify structural uncertainty to have confidence in our inference and predictions. In this work, we do so by constructing an approximate posterior distribution over CRN structures. This is done by keeping a large set of suboptimal solutions found in an optimisation framework with sparse regularisation, in contrast to existing optimisation approaches which discard suboptimal solutions. We find that inducing reaction sparsity with nonconvex penalty functions results in more parsimonious CRNs compared to the popular lasso regularisation. In a real-data example where multiple CRNs have been previously suggested, we simultaneously recover reactions proposed from different literature under our approach to structural uncertainty. We demonstrate how posterior correlations between reactions help identify where structural ambiguities are present. This can be translated into alternative reaction pathways suggested by the available data, thus guiding the efforts of future experimental design.