Single-cell RNA-seq denoising using a deep count autoencoder

Nat Commun. 2019 Jan 23;10(1):390. doi: 10.1038/s41467-018-07931-2.


Single-cell RNA sequencing (scRNA-seq) has enabled researchers to study gene expression at a cellular resolution. However, noise due to amplification and dropout may obstruct analyses, so scalable denoising methods for increasingly large but sparse scRNA-seq data are needed. We propose a deep count autoencoder network (DCA) to denoise scRNA-seq datasets. DCA takes the count distribution, overdispersion and sparsity of the data into account using a negative binomial noise model with or without zero-inflation, and nonlinear gene-gene dependencies are captured. Our method scales linearly with the number of cells and can, therefore, be applied to datasets of millions of cells. We demonstrate that DCA denoising improves a diverse set of typical scRNA-seq data analyses using simulated and real datasets. DCA outperforms existing methods for data imputation in quality and speed, enhancing biological discovery.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Animals
  • Blood Cells
  • Caenorhabditis elegans / genetics
  • Computational Biology / methods*
  • Gene Expression Profiling / methods*
  • Gene Expression Regulation / genetics
  • Leukocytes, Mononuclear
  • Models, Statistical
  • Phenotype
  • RNA / analysis
  • RNA / genetics*
  • RNA, Small Cytoplasmic / genetics
  • Sequence Analysis, RNA / methods*
  • Single-Cell Analysis / methods


  • RNA, Small Cytoplasmic
  • RNA