Illustration of the ARTIVA procedure. (A) Schematic illustration of the two-step RJ-MCMC scheme used for determining the stationary distribution of time varying dynamic Bayesian networks. With probabilities b, d and v, we propose the birth, death or shift of a changepoint (CP) respectively; with probability w we propose an update of the regression model describing regulatory interactions for a gene within a temporal phase. Varying the number of CPs or the number of edges (network topology) corresponds to a change in the dimension of the state-space and is dealt with by using Green's RJ-MCMC formalism . Proposed shifts in changepoint positions are accepted according to a standard Metropolis-Hastings step. Because of conservation of probability we necessarily have b + d + v + w = 1 and χ + ζ + ρ = 1. (B) Outline of the ARTIVA inference procedure.