A unified statistical framework for single cell and bulk RNA sequencing data

Abstract

Recent advances in technology have enabled the measurement of RNA levels for individual cells. Compared to traditional tissue-level bulk RNA-seq data, single cell sequencing yields valuable insights about gene expression profiles for different cell types, which is potentially critical for understanding many complex human diseases. However, developing quantitative tools for such data remains challenging because of high levels of technical noise, especially the “dropout” events. A “dropout” happens when the RNA for a gene fails to be amplified prior to sequencing, producing a “false” zero in the observed data. In this paper, we propose a Unified RNA-Sequencing Model (URSM) for both single cell and bulk RNA-seq data, formulated as a hierarchical model. URSM borrows the strength from both data sources and carefully models the dropouts in single cell data, leading to a more accurate estimation of cell type specific gene expression profile. In addition, URSM naturally provides inference on the dropout entries in single cell data that need to be imputed for downstream analyses, as well as the mixing proportions of different cell types in bulk samples. We adopt an empirical Bayes’ approach, where parameters are estimated using the EM algorithm and approximate inference is obtained by Gibbs sampling. Simulation results illustrate that URSM outperforms existing approaches both in correcting for dropouts in single cell data, as well as in deconvolving bulk samples. We also demonstrate an application to gene expression data on fetal brains, where our model successfully imputes the dropout genes and reveals cell type specific expression patterns.

Droplet scRNA-seq is not zero-inflated

“…On a computer science oriented branch of the scRNA-seq methods field, many methods have been designed to correct dropout zeros in data, with the aim of letting a user predict what the expression level of a gene in a cell would have been, had there been no zero-inflation or dropouts (Azizi et al, 2017;van Dijk et al, 2018;Gong et al, 2018;Huang et al, 2018;Li and Li, 2018;Tang et al, 2018;Zhu et al, 2016).…”

Separating measurement and expression models clarifies confusion in single-cell RNA sequencing analysis

“…The term "dropout" has become commonly used in connection with the zeros in scRNAseq data [5][6][7][8][9] . Historically, dropout referred to a particular failure mode of PCR, allelic dropout, in which specific primers would fail to amplify sequences containing a particular allele of a polymorphism, leading to genotyping errors for heterozygous individuals 10,11 .…”

“…The term "missing data" is also commonly used in connection with zeros in scRNAseq 8,12 . This terminology is misleading because the zeros are not missing data, at least not in the way this term is conventionally used in statistics.…”

“…Unfortunately, inappropriate use of the term "missing data" has lead researchers to apply inappropriate statistical methods for dealing with missing data to scRNA-seq data. For example, in other statistical applications it is common to attempt to "impute" (fill in) missing values, and so viewing the zeros in scRNA-seq as missing has led to many papers describing methods to "impute" the zeros 8,[13][14][15][16][17] . However, since the zeros in RNA-seq are not actually missing data, the meaning of these imputed values is unclear, and varies from method to method.…”

netNMF-sc: Leveraging gene-gene interactions for imputation and dimensionality reduction in single-cell expression analysis

“…After simulating transcript counts to obtain a complete count matrix X , we generated dropout events using one of two models. The first is a multinomial dropout model, used previously to model dropout in scRNA-seq data Linderman et al [2018], Zhu et al [2018]. In this model, the observed counts in each cell are multinomial distributed, where the probability of observing a transcript from gene i in cell j is x i j ∑ r,s x rs and the number of trials is the sum of all transcripts in the complete count matrix, ∑ r,s x rs , multiplied by the capture efficiency, ranging from 0 to 1.…”