Zero-preserving imputation of single-cell RNA-seq data

• Published in: Nature communications Vol. 13; no. 1; pp. 192 - 11
• Main Authors: Linderman, George C., Zhao, Jun, Roulis, Manolis, Bielecki, Piotr, Flavell, Richard A., Nadler, Boaz, Kluger, Yuval
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 11.01.2022; Nature Publishing Group; Nature Portfolio
• Subjects: 631/114/2415; 631/208/514/1949; Algorithms; Animals; Approximation; B-Lymphocytes - cytology; B-Lymphocytes - metabolism; Bronchi - cytology; Bronchi - metabolism; Datasets; Datasets as Topic; Epithelial Cells - cytology; Epithelial Cells - metabolism; Gene sequencing; Genes; Humanities and Social Sciences; Humans; Killer Cells, Natural - cytology; Killer Cells, Natural - metabolism; Mathematical analysis; Mice; Monocytes - cytology; Monocytes - metabolism; multidisciplinary; Primary Cell Culture; Ribonucleic acid; RNA; RNA - genetics; RNA - metabolism; RNA-Seq; Science; Science (multidisciplinary); Sequence Analysis, RNA - statistics & numerical data; Single-Cell Analysis; T-Lymphocytes - cytology; T-Lymphocytes - metabolism
• ISSN: 2041-1723; 2041-1723
• DOI: 10.1038/s41467-021-27729-z

A key challenge in analyzing single cell RNA-sequencing data is the large number of false zeros, where genes actually expressed in a given cell are incorrectly measured as unexpressed. We present a method based on low-rank matrix approximation which imputes these values while preserving biologically non-expressed genes (true biological zeros) at zero expression levels. We provide theoretical justification for this denoising approach and demonstrate its advantages relative to other methods on simulated and biological datasets. Missing values in scRNA-seq datasets can bias their analysis. Here, the authors threshold the low rank approximation of the expression matrix, so false zeros can be imputed while true zeros are preserved.