B0424
Title: Gibbs-type random partition aligned on a graph: Application to single-cell RNA data
Authors: Giovanni Rebaudo - University of Turin and Collegio Carlo Alberto (Italy) [presenting]
Peter Mueller - UT Austin (United States)
Abstract: Bayesian nonparametric mixtures and random partition models are effective tools to perform probabilistic clustering. However, standard independent mixture models can be restrictive in some applications, such as inference on cell-lineage due to the biological relations of the clusters. Motivated by single cells RNA application, we develop a novel dependent mixture model to jointly perform cluster analysis and align the cluster on a graph. Our flexible random partition model aligned on a graph cleverly exploits Gibbs type random partition as building blocks allowing suitable variable augmentations to derive analytical results on the partition distribution. From the law of the random partition, we derive a generalisation of the well-known Chinese restaurant process and a related simple MCMC algorithm to perform Bayesian inference. We illustrate the effectiveness of our proposal both on synthetic data and RNA expressions of stem cells.
