Effective and scalable single-cell data alignment with non-linear canonical correlation analysis

Hu, Jialu; Chen, Mengjie; Zhou, Xiang

doi:10.1093/nar/gkab1147

Cited by 16 publications

(15 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The rapid accumulation of large-scale single-cell datasets requires integration algorithms to efficiently handle datasets containing millions of cells without loss of accuracy. For a comprehensive comparison, we first benchmarked Portal and existing representative methods, including Harmony [21], Seurat v3 [22], online iNMF [23], VIPCCA [24], scVI [25], fastMNN [26], Scanorama [27] and BBKNN [28], in terms of integration performance following a recent benchmarking study [30]. Using a number of scRNA-seq datasets from diverse tissue types with curated cell cluster annotations, including mouse spleen, marrow, and bladder [7], we quantitatively evaluated the integration performance of each method.…”

Section: Resultsmentioning

confidence: 99%

“…scANVI [71] is another VAE-based method with similar pros and cons of scVI, as it is an extension of scVI that incorporates cell type information into its model. Recently, VIPCCA [24] was proposed to leverage VAE-based networks to perform nonlinear canonical correlation analysis (CCA) efficiently. However, we empirically found that it favors the removal of batch effects over the preservation of biological information (Figs.…”

Section: Discussionmentioning

confidence: 99%

“…Many methods have been developed to align single-cell datasets [10], including Harmony [21], Seurat [22], online iNMF [23], VIPCCA [24], scVI [25], fastMNN [26], Scanorama [27] and BBKNN [28]. Several of these methods that were designed for large datasets at the time of publication are now less attractive in terms of scalability in the face of atlas-level dataset sizes.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Adversarial domain translation networks for fast and accurate integration of large-scale atlas-level single-cell datasets

Zhao

Wang

Ming

et al. 2021

Preprint

View full text Add to dashboard Cite

The rapid emergence of large-scale atlas-level single-cell RNA-sequencing (scRNA-seq) datasets from various sources presents remarkable opportunities for broad and deep biological investigations through integrative analyses. However, harmonizing such datasets requires integration approaches to be not only computationally scalable, but also capable of preserving a wide range of fine-grained cell populations. We created Portal, a unified framework of adversarial domain translation to learn harmonized representations of datasets. With innovation in model and algorithm designs, Portal achieves superior performance in preserving biological variation during integration, while having significantly reduced running time and memory compared to existing approaches, achieving integration of millions of cells in minutes with low memory consumption. We demonstrate the efficiency and accuracy of Portal using diverse datasets ranging from mouse brain atlas projects, the Tabula Muris project, and the Tabula Microcebus project. Portal has broad applicability and in addition to integrating multiple scRNA-seq datasets, it can also integrate scRNA-seq with single-nucleus RNA-sequencing (snRNA-seq) data. Finally, we demonstrate the utility of Portal by applying it to the integration of cross-species datasets with limited shared-information between them, and are able to elucidate biological insights into the similarities and divergences in the spermatogenesis process between mouse, macaque, and human.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Adversarial domain translation networks for fast and accurate integration of large-scale atlas-level single-cell datasets

Zhao

Wang

Ming

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Second, the fast-evolving technology of single-cell omics provides opportunities to integrate omics profiles from different modalities for the same individuals. Extending DR-SC by integrating multiple different omics techniques, such as through the canonical correlation analysis framework [80] which is a nature extension of PCA towards multiple modality analysis, will also likely achieve higher statistical performance. Third, DR-SC essentially performs unsupervised clustering, but with the availability of labels for some cells/spots, it would be interesting to perform semi-supervised clustering of those data.…”

Section: Discussionmentioning

confidence: 99%

Joint dimension reduction and clustering analysis for single-cell RNA-seq and spatial transcriptomics data

Liu

Liao

Zhou

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Dimension reduction and (spatial) clustering are two key steps for the analysis of both single-cell RNA-sequencing (scRNA-seq) and spatial transcriptomics data collected from different platforms. Most existing methods perform dimension reduction and (spatial) clustering sequentially, treating them as two consecutive stages in tandem analysis. However, the low-dimensional embeddings estimated in the dimension reduction step may not necessarily be relevant to the class labels inferred in the clustering step and thus may impair the performance of the clustering and other downstream analysis. Here, we develop a computation method, DR-SC, to perform both dimension reduction and (spatial) clustering jointly in a unified framework. Joint analysis in DR-SC ensures accurate (spatial) clustering results and effective extraction of biologically informative low-dimensional features. Importantly, DR-SC is not only applicable for cell type clustering in scRNA-seq studies but also applicable for spatial clustering in spatial transcriptimics that characterizes the spatial organization of the tissue by segregating it into multiple tissue structures. For spatial transcriptoimcs analysis, DR-SC relies on an underlying latent hidden Markov random field model to encourage the spatial smoothness of the detected spatial cluster boundaries. We also develop an efficient expectation-maximization algorithm based on an iterative conditional mode. DR-SC is not only scalable to large sample sizes, but is also capable of optimizing the spatial smoothness parameter in a data-driven manner. Comprehensive simulations show that DR-SC outperforms existing clustering methods such as Seurat and spatial clustering methods such as BayesSpace and SpaGCN and extracts more biologically relevant features compared to the conventional dimension reduction methods such as PCA and scVI. Using 16 benchmark scRNA-seq datasets, we demonstrate that the low-dimensional embeddings and class labels estimated from DR-SC lead to improved trajectory inference. In addition, analyzing three published scRNA-seq and spatial transcriptomics data in three platforms, we show DR-SC can improve both the spatial and non-spatial clustering performance, resolving a low-dimensional representation with improved visualization, and facilitate the downstream analysis such as trajectory inference.

show abstract

“…Sequence compositions of nucleic acids and proteins are significantly associated with genome evolution and adaptation across all kingdoms of life [ 17 ]. Machine/deep learning methods have worked well in predicting viral hosts based on the amino acid [ 18 ] or dinucleotide (DNT) [ 19 ] composition in the sequence alignment of large datasets [ 20 ]. Moreover, language representation methods have learned the language of viral evolution and escape based on represented amino acids [ 21 ] or statistically represented proteins [ 22 ].…”

Section: Introductionmentioning

confidence: 99%

Deep learning based on biologically interpretable genome representation predicts two types of human adaptation of SARS-CoV-2 variants

Zhang

et al. 2022

Briefings in Bioinformatics

View full text Add to dashboard Cite

Explosively emerging SARS-CoV-2 variants challenge current nomenclature schemes based on genetic diversity and biological significance. Genomic composition-based machine learning methods have recently performed well in identifying phenotype–genotype relationships. We introduced a framework involving dinucleotide (DNT) composition representation (DCR) to parse the general human adaptation of RNA viruses and applied a three-dimensional convolutional neural network (3D CNN) analysis to learn the human adaptation of other existing coronaviruses (CoVs) and predict the adaptation of SARS-CoV-2 variants of concern (VOCs). A markedly separable, linear DCR distribution was observed in two major genes—receptor-binding glycoprotein and RNA-dependent RNA polymerase (RdRp)—of six families of single-stranded (ssRNA) viruses. Additionally, there was a general host-specific distribution of both the spike proteins and RdRps of CoVs. The 3D CNN based on spike DCR predicted a dominant type II adaptation of most Beta, Delta and Omicron VOCs, with high transmissibility and low pathogenicity. Type I adaptation with opposite transmissibility and pathogenicity was predicted for SARS-CoV-2 Alpha VOCs (77%) and Kappa variants of interest (58%). The identified adaptive determinants included D1118H and A570D mutations and local DNTs. Thus, the 3D CNN model based on DCR features predicts SARS-CoV-2, a major type II human adaptation and is qualified to predict variant adaptation in real time, facilitating the risk-assessment of emerging SARS-CoV-2 variants and COVID-19 control.

show abstract

Effective and scalable single-cell data alignment with non-linear canonical correlation analysis

Cited by 16 publications

References 51 publications

Adversarial domain translation networks for fast and accurate integration of large-scale atlas-level single-cell datasets

Adversarial domain translation networks for fast and accurate integration of large-scale atlas-level single-cell datasets

Joint dimension reduction and clustering analysis for single-cell RNA-seq and spatial transcriptomics data

Deep learning based on biologically interpretable genome representation predicts two types of human adaptation of SARS-CoV-2 variants

Contact Info

Product

Resources

About