Bayesian Hierarchical Model for Single Cell Hi-C Matrix Imputation and Its Impact on Chromosome 3D Structure

• Activity Number: 81 - Contributed Poster Presentations: Section on Statistics in Epidemiology
• Type: Contributed
• Date/Time: Monday, August 3, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Qing Xie* and Chenggong Han and Shili Lin
• Companies: The Ohio State University and Ohio State University and Ohio State University
• Keywords: Single cell Hi-C; RNA-seq; imputation; 3D structure; Bayesian hierarchy; MCMC

Abstract

The prevalence of dropout events has been a huge problem for single cell Hi-C data, which brings in many difficulties in downstream studies such as clustering and structural analysis. Although numerous efforts have been spent on imputing dropout events for single cell RNA-seq data, there has been little research on imputing single cell Hi-C data. In this paper, we proposed an approach by which single cell RNA-seq imputation techniques can be applied directly to single cell Hi-C data, for both multiple single cells scenario and one single cell scenario. Besides, we developed a Bayesian hierarchical model to distinguish the structural zero from the sampling zero for single cell Hi-C data and make imputations at the sampling zero positions, taking bulk cells Hi-C into consideration. Simulations were performed to compare the proposed method and several existing imputation methods. Follow-up analysis shows that the clustering and the construction of 3D structure of single cells by their Hi-C data has been improved after imputation.