A Comparison for Dimensionality Reduction Methods of Single-Cell RNA-seq Data

Abstract

Single-cell RNA sequencing (scRNA-seq) is a high-throughput sequencing technology performed at the level of an individual cell, which can have a potential to understand cellular heterogeneity. However, scRNA-seq data are high-dimensional, noisy, and sparse data. Dimension reduction is an important step in downstream analysis of scRNA-seq. Therefore, several dimension reduction methods have been developed. We developed a strategy to evaluate the stability, accuracy, and computing cost of 10 dimensionality reduction methods using 30 simulation datasets and five real datasets. Additionally, we investigated the sensitivity of all the methods to hyperparameter tuning and gave users appropriate suggestions. We found that t-distributed stochastic neighbor embedding (t-SNE) yielded the best overall performance with the highest accuracy and computing cost. Meanwhile, uniform manifold approximation and projection (UMAP) exhibited the highest stability, as well as moderate accuracy and the second highest computing cost. UMAP well preserves the original cohesion and separation of cell populations. In addition, it is worth noting that users need to set the hyperparameters according to the specific situation before using the dimensionality reduction methods based on non-linear model and neural network.

Deciphering the Retinal Epigenome during Development, Disease and Reprogramming: Advancements, Challenges and Perspectives

“…A feature matrix can be built upon signals originating from individual cells across genomic coordinates and the deriving, aggregated signal can then be subject to unsupervised clustering by k-means, Louvain, or hierarchical clustering [ 369 ]. The signals are then subject to dimensionality reduction and visualization by principal component analysis (PCA), t-SNE or projected with a uniform manifold learning technique for dimensionality reduction (UMAP) [ 370 , 371 ] to identify putative subpopulations and similarities across individual profiles [ 372 , 373 , 374 ]. The quality of clustering can then be assessed by comparison with bulk sequencing from flow-sorted cells purified by virtue of a reporter or based on the expression of known marker genes.…”

Integration of Machine Learning and Coarse-Grained Molecular Simulations for Polymer Materials: Physical Understandings and Molecular Design

“…Besides, the permutation invariance is enforced by performing the Gaussian expansion and pooling to yield a histogram of features . This histogram was then reshaped and embedded into a low-dimensional latent space Z 0 using a Uniform Manifold Approximation and Projection (UMAP) approach, which is a non-linear, unsupervised method for dimension reduction ( Xiang et al, 2021 ) . By doing so, a projection of was obtained in the latent space Z 0 , and then local structural information of monomers was achieved based on their positions in this manifold .…”

Analyzing Modern Biomolecules: The Revolution of Nucleic-Acid Sequencing – Review

“…Therefore, a combination of several of them may be needed for a comprehensive analysis of gene expression [125]. Besides, computational models [126], such as ML, have been applied to these studies [127], including dimension reduction methods [128]. Bioinformatics developments have also allowed to deconvult heterogeneous cell samples [129], as well as identify pathways or biological processes from transcriptomics [130].…”