Singletrome: A method to analyze and enhance the transcriptome with long noncoding RNAs for single cell analysis

Abstract

AbstractSingle cell RNA sequencing (scRNA-seq) has revolutionized the study of gene expression in individual cell types from heterogeneous tissue. To date, scRNA-seq studies have focused primarily on expression of protein-coding genes, as the functions of these genes are more broadly understood and more readily linked to phenotype. However, long noncoding RNAs (lncRNAs) are even more diverse than protein-coding genes, yet remain an underexplored component of scRNA-seq data. While less is known about lncRNAs, they are widely expressed and regulate cell development and the progression of diseases including cancer and liver disease. Dedicated lncRNA annotation databases continue to expand, but most lncRNA genes are not yet included in reference annotations applied to scRNA-seq analysis. Simply creating a new annotation containing known protein-coding and lncRNA genes is not sufficient, because the addition of lncRNA genes that overlap in sense and antisense with protein-coding genes will affect how reads are counted for both protein-coding and lncRNA genes. Here we introduce Singletrome, an enhanced human lncRNA genome annotation for scRNA-seq analysis, by merging protein-coding and lncRNA databases with additional filters for quality control. Using Singletrome to characterize expression of lncRNAs in human peripheral blood mononuclear cell (PBMC) and liver scRNA-seq samples, we observed an increase in the number of reads mapped to exons, detected more lncRNA genes, and observed a decrease in uniquely mapped transcriptome reads, indicating improved mapping specificity. Moreover, we were able to cluster cell types based solely on lncRNAs expression, providing evidence of the depth and diversity of lncRNA reads contained in scRNA-seq data. Our analysis identified lncRNAs differentially expressed in specific cell types with development of liver fibrosis. Importantly, lncRNAs alone were able to predict cell types and human disease pathology through the application of machine learning. This comprehensive annotation will allow mapping of lncRNA expression across cell types of the human body facilitating the development of an atlas of human lncRNAs in health and disease.