CellHarmony: Cell-level matching and comparison of single-cell transcriptomes

Abstract

To understand the molecular pathogenesis of human disease, precision analyses to define molecular alterations within (and between) disease-associated cell populations are desperately needed. Single-cell genomics represents an ideal platform to enable the identification and comparison of normal and diseased transcriptional cell states. We note that disease-associated perturbations usually retain cellular-identity programs (core genes), providing an appropriate reference for secondary comparison analyses. Thus, we created cellHarmony, an integrated solution for the unsupervised analysis and classification of cell types from diverse scRNA-Seq datasets. cellHarmony is an automated and easy-to-use tool that efficiently matches single-cell transcriptomes using a community clustering and alignment strategy. Utilizing core genes and community clustering to reveal disease and cell-state systems-level insights overcomes bias toward donor and disease effects that can be imposed by joint-alignment approaches. Moreover, cellHarmony directly compares cell frequencies and gene expression in a cell-type-specific manner, then produces a holistic representation of these differences across potentially dozens of cell populations and impacted regulatory networks. Using this approach, we identify gene regulatory programs that are selectively impacted in distinct hematopoietic and heart cell populations that suggest novel disease mechanisms and drug targets. Thus, this approach holds tremendous promise in revealing the molecular and cellular origins of complex diseases.