Skip to main content

Table 1 List of compared dimensionality reduction methods. We list standard modeling properties for each of compared dimensionality reduction methods

From: Accuracy, robustness and scalability of dimensionality reduction methods for single-cell RNA-seq analysis

No.MethodsModeling countsModeling zero inflationNon-linear projectionComputation efficiencyImplementation languageYear of publicationReference
1PCANoNoNoYesR1901[42]
2ICANoNoNoNoR1994[43]
3FANoNoNoYesR1952[44]
4NMFNoNoNoYesR1999[45]
5Poisson NMFYesNoNoYesR1999[45]
6Diffusion MapNoNoYesYesR2005[46]
7ZIFANoYesNoNoPython2016[30]
8ZINB-WaVEYesYesNoNoR2018[32]
9GLMPCAYesNoNoNoR2019[47]
10pCMFYesYesNoNoR2019[31]
11scScopeNoYesYesYesPython2019[37]
12DCAYesYesYesYesPython2018[40]
13tSNENoNoYesNoR2008[48]
14MDSNoNoNoYesR1958[49]
15LLENoNoYesYesR2000[50]
16LTSANoNoYesNoR2004[51]
17IsomapNoNoYesYesR2000[11]
18UMAPNoNoYesYesPython2019[52]
  1. These properties include whether it models count data (3rd column), whether it accounts for zero inflation (4th column), whether it is a linear dimensionality reduction method (5th column), its computation efficiency (6th column), implementation language (7th column), year of publication (8th column), and reference (9th column). FA factor analysis, PCA principal component analysis, ICA independent component analysis, NMF nonnegative matrix factorization, Poisson NMF Kullback-Leibler divergence-based NMF, ZIFA zero-inflated factor analysis, ZINB-WaVE zero-inflated negative binomial-based wanted variation extraction, pCMF probabilistic count matrix factorization, DCA deep count autoencoder network, scScope scalable deep-learning-based approach, GLMPCA generalized linear model principal component analysis, Diffusion Map, MDS multidimensional scaling, LLE locally linear embedding, LTSA local tangent space alignment, Isomap; UMAP uniform manifold approximation and projection, tSNE t-distributed stochastic neighbor embedding