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

By Gökçen Eraslan, Lukas M. Simon, Maria-Madalina Mircea, N. Mueller, Fabian J Theis. Nature Communications, 2018

996 citations recorded by the source corpus.

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. Single-cell RNA sequencing is a powerful method to study gene expression, but noise in the data can obstruct analysis. Here the authors develop a denoising method based on a deep count autoencoder network that scales linearly with the number of cells, and therefore is compatible with large data sets.

Subfields: Molecular Biology

View Paper: https://www.semanticscholar.org/paper/d9c208af563534ed6add7b8e1e9372909e850f83

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

Citations in the corpus

Bayesian Inference for a Generative Model of Transcriptome Profiles from Single-cell RNA Sequencing (2018, methodology): “DCA [27] is a denoising auto-encoder [38] that proposes a ZINB loss function.”
Graph abstraction reconciles clustering with trajectory inference through a topology preserving map of single cells (2017, background): “Using (46), we can now quantify these violations and show that PAGA-initialized manifold learning both provides embeddings that are more faithful to the global topology and allows faster convergence also with respect to the conventional cost function (43).”
Droplet scRNA-seq is not zero-inflated (2019, methodology): “On a statistics oriented branch of the scRNA-seq methods field, the zero-inflated negative binomial distribution have become popular as a way to directly model scRNA-seq data (Eraslan et al., 2018; Lopez et al., 2018; Risso et al., 2018).”
Clustering single-cell RNA-seq data with a model-based deep learning approach (2019, methodology): “The results of DCA also show that its imputation using ZINB can improve a diverse set of typical downstream scRNA-seq data analyses 23 .”
Predicting cellular responses to complex perturbations in high‐throughput screens (2023, methodology): “Although we opted to implement a deterministic autoencoder scheme, extensions toward variational models (Lopez et al , 2018; Lotfollahi et al , 2020), as well as cost functions other than mean squared error (Eraslan et al , 2019) are straightforward.”
Spatial components of molecular tissue biology (2022, positive): “A causally interpretable model of these extrinsic effects in the microenvironment requires dependencies between cells and violates the independence assumption commonly used in cell-centric (intrinsic) models of gene expression in many analyses of scRNA-seq 37–39 .”
Probabilistic harmonization and annotation of single‐cell transcriptomics data with deep generative models (2019, neutral): “This concern has been addressed to some extent by recent progress in the field, proposing search algorithms based on held-out log-likelihood maximization [59].”
Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression (2019, methodology): “Importantly, statistical and deep-learning methods designed for single-cell RNA-seq data often utilize a negative binomial (or zero-inflated negative binomial) error model [10, 15].”
Unsupervised removal of systematic background noise from droplet-based single-cell experiments using CellBender (2019, result): “The current opinion in biology, as well as previous experiments with neural auto-encoders in the literature [12, 13], agree with the manifold embedding hypothesis.”
Current best practices in single‐cell RNA‐seq analysis: a tutorial (2019, methodology): “First applications to scRNA-seq are starting to emerge from dimensionality reduction to denoising (e.g. scVis: Ding et al , 2018; scGen: preprint: Lotfollahi et al , 2018; DCA: Eraslan et al , 2019).”
Deep Generative Modeling for Single-cell Transcriptomics (2018, methodology): “We also observe similar levels of scalability with DCA [20] a denoising autoencoder, which also uses stochastic optimization.”