PBLR: an accurate single cell RNA-seq data imputation tool considering cell heterogeneity and prior expression level of dropouts

Abstract

Single-cell RNA sequencing (scRNA-seq) provides a powerful tool to determine precise expression patterns of tens of thousands of individual cells, decipher cell heterogeneity and cell subpopulations and so on. However, scRNA-seq data analysis remains challenging due to various technical noise, e.g., the presence of dropout events (i.e., excess zero counts). Taking account of cell heterogeneity and structural effect of expression on dropout rate, we propose a novel method named PBLR to accurately impute the dropouts of scRNA-seq data. PBLR is an effective tool to recover dropout events on both simulated and real scRNA-seq datasets, and can dramatically improve low-dimensional representation and recovery of gene-gene relationship masked by dropout events compared to several state-of-the-art methods. Moreover, PBLR also detect accurate and robust cell subpopulations automatically, shedding light its flexibility and generality for scRNA-seq data analysis.