Bayesian non-parametric methods are a natural approach to fitting models with continuous parameters or unbounded parameter set cardinality. In our group we are applying these methods to diverse models of time-series gene expression data. Example applications include differential equation models of transcriptional regulation, clustering data sampled at uneven times and phylogenetic models of gene expression change over evolutionary time. We use continuous-time Gaussian processes to model the time-evolution of gene expression and protein activation/concentration in time. Dirichlet processes can also be used to model an unbounded set of Gaussian process models. I will present results where we apply these methods to gene expression time-course data from embryonic development in Drosophila.