Bayesian Nonparametric Inference of Effective Population Trajectories with Gaussian Processes

• Activity Number: 360
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Julia Adela Palacios*+ and Vladimir Minin
• Companies: University of Washington and University of Washington
• Keywords: effective population size ; coalescent ; Sigmoidal Gaussian Cox Process ; MCMC

Abstract

Our goal is Bayesian inference of effective population size trajectories from conteporaneously or serially sampled molecular sequences under the coalescent. We put a sigmoidal Gaussian process prior on effective population size trajectories without assuming a parametric form of these trajectories. We adapt recent advances in MCMC algorithms for sigmoidal Gaussian Cox processes to approximate the posterior distribution of all coalescent model parameters. Importantly, our MCMC algorithm is exact in that it does not require finite-dimensional approximations of the Gaussian process. We validate our method using simulated and real data and compare it to competing Gaussian Markov random field smoothing and change-point model methods.