A new bioinformatics tool to recover missing gene expression in single-cell RNA sequencing data

Single-cell RNA sequencing (scRNA-seq) is a burgeoning field where experimental techniques and computational methods have been under rapid evolution in the past six years. These technological advances have allowed biomedical researchers to identify new cell types, delineate cell sub-populations, and infer cell differentiation trajectories in various tissue samples. Among the important features extractable from scRNA-seq data, the predominant ones are individual genes' expression levels in single cells. Most analyses require a preprocessing step that converts a scRNA-seq dataset into a count matrix, where rows correspond to cells (or genes), columns correspond to genes (or cells), and entries are counts, i.e. a count is the number of sequenced reads or uniquely mapped identifiers (UMIs) mapped to a gene in a cell. Single-cell count matrices are highly sparse; for example, a typical matrix constructed from a droplet-based dataset may have >90% of counts as zeros.