A Bayesian mixture model for clustering droplet-based single-cell transcriptomic data from population studies.

Area of Science:

• Computational Biology
• Genomics
• Statistical Genetics

Background:

• Droplet-based single-cell transcriptome sequencing (scRNA-seq) enables population-scale studies, measuring transcriptomes in tens of thousands of single cells across individuals.
• Existing clustering methods often lack specific adaptations for the complexities of population-scale scRNA-seq data, particularly concerning inter-individual variability.

Purpose of the Study:

• To develop a novel computational method, BAMM-SC (Bayesian Analysis of Mixture Models for Single-Cell sequencing), for simultaneously clustering scRNA-seq data from multiple individuals.
• To address data heterogeneity and batch effects inherent in population-scale, multi-individual scRNA-seq studies within a unified statistical framework.

Main Methods:

• Developed BAMM-SC, a Bayesian mixture model designed to process raw count data from scRNA-seq.
• The model employs a unified Bayesian hierarchical framework to simultaneously analyze data from multiple individuals.
• Incorporates specific mechanisms to account for heterogeneity and batch effects among different individuals.

Main Results:

• Extensive simulation studies demonstrated BAMM-SC's superior performance compared to existing clustering methods.
• Application to human and mouse blood, lung, and skin cell datasets confirmed improved clustering accuracy.
• BAMM-SC showed particular effectiveness in handling datasets with significant inter-individual heterogeneity.

Conclusions:

• BAMM-SC provides a robust and accurate method for clustering population-scale scRNA-seq data from multiple individuals.
• The method effectively mitigates challenges posed by data heterogeneity and batch effects, enhancing biological insights.
• BAMM-SC represents a significant advancement for analyzing large-scale single-cell genomics studies.