Bayesian model discovery for reverse-engineering biochemical networks from data

Reverse engineering gene regulatory networks from gene expression data is a challenging inference task. A related problem in computational systems biology is identification of signalling networks that perform particular functions, such as adaptation. Indeed, for many research questions, there is an ongoing need for efficient inference algorithms that can identify the simplest model, from among a larger set of inter-related models, that best explains empirical observations. To this end, we introduce Sparse Likelihood-free Inference using Gibbs sampling (SLInG), a Bayesian sparse likelihood-free inference method. SLInG provides an efficient sampling method for Approximate Bayesian Computation with sparsity-inducing hierarchical priors that is widely applicable for any simulation-based model discovery task. We first apply SLInG to linear sparse regression problem using a classic dataset, before focusing on applications to biochemical network model discovery. We demonstrate that SLInG can reverse engineer stochastic gene regulatory networks from single-cell data with high accuracy, outperforming state-of-the-art correlation-based methods. Furthermore, we show that SLInG can successfully identify signalling networks that execute adaptation. Sparse hierarchical Bayesian inference thus provides a versatile and powerful tool for model discovery in systems biology and beyond.