SAME-Clustering: Single-Cell Aggregated Clustering via Mixture Model Ensemble

• Activity Number: 217 - Studying Psychiatric Disorders Using Statistical and Machine Learning Methods
• Type: Invited
• Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Mental Health Statistics Section
• Author(s): Ruth Huh and Yuchen Yang and Houston Culpepper and Yun Li*
• Companies: University of North Carolina at Chapel Hill and University of North Carolina at Chapel Hill and University of North Carolina at Chapel Hill and University of North Carolina at Chapel Hill
• Keywords: scRNA-seq; cluster ensemble; multinomial mixture; mixture model

Abstract

Clustering single-cell RNA-seq (scRNA-seq) data is an important task. Clustering results themselves are of great importance for shedding light on tissue complexity. Inferred cell types from clustering analysis are valuable for differential expression analysis where we need to adjust for them to mitigate deceiving results. Several novel methods have been developed for clustering scRNA-seq data. However, different approaches generate varying cluster assignments and number of clusters. It is usually hard to gauge which method to use because none of the clustering methods always outperform the other methods in all datasets. SAME-clustering adopts a probabilistic model to build a consensus by using a finite mixture model of multinomial distributions. Our current implementation takes clustering results from four methods, SC3, CIDR, Seurat, and t-SNE+kmeans, as input and produces ensemble results using the EM algorithm. We have tested SAME-clustering across 14 datasets and results show that our method yields enhanced cluster results. Our applications to scRNA-seq data in brain tissues shed light on genetic mechanisms underlying psychiatric disorders.