Single-Cell Gene Expression Variability: Functional Assessment and Applications in Functional Genomics

Abstract

Gene expression variability has been associated with specific roles in cell function. However, its functional implications in multicellular organization have not been systematically tested due to technical limitations. Furthermore, their potential application in functional genomics that may increase our understanding of the regulatory mechanisms driving different cell states and the active role of the genes on them have not been evaluated. Thanks to the development of single-cell RNA-seq techniques allowing the measurement of the transcriptome profile in thousands of cells in a single experiment, now it is possible to characterize the synchronized patterns of expression of genes participating in the same biological processes or under the regulation of the same transcription factor. This allows identifying cells under the same cellular state and genes’ functional relationships driving those cellular states without the need for genetic manipulations.

Here we introduce three new single-cell RNA-seq datasets (from lymphoblastoid cell lines, Ahr, and Malat1 knockouts). We also introduce cell-type and tissue-specific thresholds for single-cell RNA-seq quality control and novel computational methods to increase our understanding of the biological implications of the single-cell gene expression variability in a high-level order in multicellular organisms and its applications for the identification of differentially regulated genes driving the observed cellular states, and the prediction of the cell-type-specific functional roles of the genes. Our results provide evidence supporting the 'variation is function' hypothesis suggesting that the aggregate cellular function may depend on the single-cell gene expression variability observed among cells of the same type under the same environment. We show that the reported thresholds for single-cell RNA-seq quality control accurately discriminate between healthy and low-quality cells in different tissues and cell-types. We also demonstrate that our novel computational methods based on gene expression variability and unsupervised machine learning algorithms allow unraveling the regulatory mechanisms underlying cell behaviors and the accurate prediction of the perturbations caused by the deletion of a gene in a gene regulatory network revealing the gene’s function in a cell type-specific manner.