%0 Journal Article %A Zhun Miao %A Xuegong Zhang %T DEsingle: A new method for single-cell differentially expressed genes detection and classification %D 2017 %R 10.1101/173997 %J bioRxiv %P 173997 %X There are excessive zero values in single-cell RNA-seq (scRNA-seq) data. Some of them are real zeros of non-expressed genes, while the others are the so-called “dropout” zeros caused by the low mRNA capture efficiency of tiny amounts of mRNAs in single cells. These two types of zeros should be distinguished in differential expression (DE) analysis and other types of analyses of scRNA-seq data. We proposed a new method DEsingle for DE analysis in scRNA-seq data by employing the Zero-Inflated Negative Binomial (ZINB) model. We proved that DEsingle could estimate the percentage of real zeros and dropout zeros by modelling the mRNA capture procedure. According to this model, DEsingle can distinguish three types of differential expression between two groups of single cells, with regard to differences in expression status, in expression abundances, and in both. We validated the performance of the method on simulation data and applied it on real scRNA-seq data of human preimplantation embryonic cells of different days of embryo development. Results showed that DEsingle outperforms existing methods for scRNA-seq DE analysis, and can reveal different types of DE genes that are enriched in different functions. %U https://www.biorxiv.org/content/biorxiv/early/2017/08/09/173997.full.pdf