ZIFA: Dimensionality reduction for zero-inflated single-cell gene expression analysis

By Emma Pierson, Christopher Yau. Genome Biology, 2015

578 citations recorded by the source corpus.

Single-cell RNA-seq data allows insight into normal cellular function and various disease states through molecular characterization of gene expression on the single cell level. Dimensionality reduction of such high-dimensional data sets is essential for visualization and analysis, but single-cell RNA-seq data are challenging for classical dimensionality-reduction methods because of the prevalence of dropout events, which lead to zero-inflated data. Here, we develop a dimensionality-reduction method, (Z)ero (I)nflated (F)actor (A)nalysis (ZIFA), which explicitly models the dropout characteristics, and show that it improves modeling accuracy on simulated and biological data sets.

Subfields: Biophysics, Molecular Biology

View Paper: https://www.semanticscholar.org/paper/4cdbfca1d466a42bdd452243b1bc748c12ca4f18

How researchers see this paper

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

In Biophysics

In Molecular Biology

Citations in the corpus

Design and computational analysis of single-cell RNA-sequencing experiments (2016, methodology): “Pierson and Yau [70] used simulation studies to show that ZIFA has comparable performance to PCA/FA when no (or few) zeros are present and has considerable advantages in the presence of zeros.”
Multi‐Omics Factor Analysis—a framework for unsupervised integration of multi‐omics data sets (2018, methodology): “…models could expand the value of MOFA in data sets with specific statistical properties that hamper the application of traditional statistical methods, including zero-inflated data (i.e. scRNA-Seq; Pierson & Yau, 2015) or binomial distributed data (i.e. splicing events; Huang & Sangui-netti, 2017).”
Modelling dropouts for feature selection in scRNASeq experiments (2016, methodology): “We compared the fit of the Michaelis-Menten model of dropouts (M3Drop) to two alternative models [3,26] across six full-transcript scRNASeq datasets (Figure 1A, S3).”
Feature selection and dimension reduction for single-cell RNA-Seq based on a multinomial model (2019, background): “For example, models based on the lognormal distribution cannot account for exact zeros, motivating the development of zero-inflated lognormal models for scRNA-Seq read counts [23, 24, 25, 26].”
Splatter: simulation of single-cell RNA sequencing data (2017, methodology): “Some analysis methods (ZIFA [11], MAST [12], ZINB-WaVE 23 [13]) incorporate dropout into their models while others (MAGIC [14], SAVER [15], 24 scImpute [16]) attempt to infer what the true expression levels should be.”
Generalizable and Scalable Visualization of Single-Cell Data Using Neural Networks (2018, background): “, dropout events where nonzero expression levels are missed as zero) have also been proposed (Pierson and Yau, 2015; Wang et al., 2017).”
Using neural networks for reducing the dimensions of single-cell RNA-Seq data (2017, methodology): “To improve its run time we reduced the number of genes to 1356 (selecting only genes that have non-zero values in 90% of samples) but the performance of ZIFA on this data was not good (much lower than the results presented in the comparison table) and so we left it out.”
DrImpute: imputing dropout events in single cell RNA sequencing data (2017, methodology): “There are some statistical tools that model dropout events for specific 318 purposes, such as BISCUIT, ZIFA, CIDR, and scImpute ([34], [35], [36], [18], 319 [14], [37]).”
Single-cell transcriptome sequencing: recent advances and remaining challenges (2016, background): “Although zero-inflated distributions can be used to accommodate expected dropout 104–106 , such models also have a greater number of parameters and can be more difficult to fit than a simpler model, particularly when sample size is limited.”
RNA Sequencing of Single Human Islet Cells Reveals Type 2 Diabetes Genes. (2016, background): “The cells segregated into distinct but homogeneous populations independent of donor type using t-distributed stochastic neighbor embedding (tSNE) (Figure 1B) or ZIFA (zero-inflated factor analysis) (Pierson and Yau, 2015) (Figure S3C).”
A deep generative model for gene expression profiles from single-cell RNA sequencing (2017, methodology): “In this case, SCONE selected a strategy that consists in scaling by depth and regressing out the QC. silhouette FA 0.208 ZIFA 0.202 ZINB-WaVE 0.260 scVI 0.285 Table 3: Silhouette scores on the mouse cortex dataset. silhouette QC correlation (higher is better) (lower is better) PCA 0.314 0.381 PCA (n”