A Bayesian Hierarchical Model for Single-Cell RNA-Sequencing Data Clustering

• Activity Number: 661 - Statistical Approaches to High-Dimensional Modeling and Real-World Problems
• Type: Contributed
• Date/Time: Thursday, August 3, 2017 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Yiyi Liu* and Hongyu Zhao
• Companies: Yale University and Yale University
• Keywords: single cell RNA-sequencing ; missing data ; Bayesian hierarchical model

Abstract

The development of single-cell RNA-sequencing (scRNA-seq) technology provides high resolution data to study the heterogeneity of cells in different tissues. However, a key challenge in using scRNA-seq is high noises and frequent dropouts (missing values) of measured RNA expression levels because there is much less input RNA than tissue-level bulk measurement that usually involves millions of cells. This noisy and zero-inflated nature of scRNA-seq data presents challenges for existing clustering methods. In this presentation, we propose a Bayesian hierarchical model that clusters cells based on scRNA-seq data through explicitly modeling dropout events. We demonstrate the effectiveness of our method with extensive simulation studies and an application to a single cell data set from a neural progenitor cell population.