scBGEDA: deep single-cell clustering analysis via a dual denoising autoencoder with bipartite graph ensemble clustering

Abstract

Motivation Single-cell RNA sequencing (scRNA-seq) is an increasingly popular technique for transcriptomic analysis of gene expression at the single-cell level. Cell-type clustering is the first crucial task in the analysis of scRNA-seq data that facilitates accurate identification of cell types and the study of the characteristics of their transcripts. Recently, several computational models based on a deep autoencoder and the ensemble clustering have been developed to analyze scRNA-seq data. However, current deep autoencoders are not sufficient to learn the latent representations of scRNA-seq data, and obtaining consensus partitions from these feature representations remains under-explored. Results To address this challenge, we propose a single-cell deep clustering model via a dual denoising autoencoder with bipartite graph ensemble clustering called scBGEDA, to identify specific cell populations in single-cell transcriptome profiles. First, a single-cell dual denoising autoencoder network is proposed to project the data into a compressed low-dimensional space and that can learn feature representation via explicit modeling of synergistic optimization of the zero-inflated negative binomial (ZINB) reconstruction loss and denoising reconstruction loss. Then, a bipartite graph ensemble clustering algorithm is designed to exploit the relationships between cells and the learned latent embedded space by means of a graph-based consensus function. Multiple comparison experiments were conducted on twenty scRNA-seq datasets from different sequencing platforms using a variety of clustering metrics. The experimental results indicated that scBGEDA outperforms other state-of-the-art methods on these datasets, and also demonstrated its scalability to large-scale scRNA-seq datasets. Moreover, scBGEDA was able to identify cell-type specific marker genes and provide functional genomic analysis by quantifying the influence of genes on cell clusters, bringing new insights into identifying cell types and characterizing the scRNA-seq data from different perspectives. Availability The source code of scBGEDA is available at https://github.com/wangyh082/scBGEDA. The software and the supporting data can be downloaded from https://figshare.com/articles/software/scBGEDA/19657911. Supplementary information Supplementary data are available at Bioinformatics online.

scGAL: unmask tumor clonal substructure by jointly analyzing independent single-cell copy number and scRNA-seq data

“…We compare scGAL to two most relevant baseline methods including Dhaka [ 20 ] and RobustClone [ 36 ] that can directly cluster single-cell copy number data, as well as three scRNA-seq based methods including Seurat [ 37 ], scBGEDA [ 38 ] and scTAG [ 39 ], which are evaluated on single-cell copy number data to check if they can conquer the inherent difference between two individual omics data and accurately identify clonal copy number substructure. Furthermore, to validate the effectiveness of the GAN module employed in scGAL for integrating standalone scRNA-seq data, an AE model that has the same structure as the AE module of scGAL and only analyzes single-cell copy number data, is used as a baseline method.…”

Recent Advancements in Subcellular Proteomics: Growing Impact of Organellar Protein Niches on the Understanding of Cell Biology

“…Raw data files generated from various subcellular proteomic methods are invaluable data resources. In MS-based subcellular proteomics studies, raw output files can be submitted to standard protein repositories: 219 PRIDE, 220 Panorama, 221 PeptideAtlas, 222 and Mass Spectrometry Interactive Virtual Environment (MassIVE). 223 For imaging-based spatial subcellular proteomics, the raw image data files can be submitted to public repositories like Image Data Resource (IDR), 224 Cell Image Library (CIL), 225 and Broad Bioimage Benchmark Collection (BC).…”

“…Raw data files generated from various subcellular proteomic methods are invaluable data resources. In MS-based subcellular proteomics studies, raw output files can be submitted to standard protein repositories: PRIDE, Panorama, PeptideAtlas, and Mass Spectrometry Interactive Virtual Environment (MassIVE) …”

scDFN: Enhancing single-cell RNA-seq Clustering with Deep Fusion Networks

“…These advanced single-cell data clustering approaches are categorized into five main groups based on their network optimization objectives. These groups include generative adversarial network-based methods (e.g., scGPCL [11] and scDECL [12]), subspace clustering-based methods (e.g., scBGEDA [13]), Gaussian mixture model-based methods (e.g., scSSA [14]), spectral clustering-based methods (e.g., Secuer [15] and scDSSC [16]), and self-optimization-based methods (e.g., scziDesk [17], scDeepCluster [18], DESC [19], and GraphSCC [20]). The methods are described in Table 1.…”

“…The scziDesk [17] model combines deep learning techniques with denoising autoencoders to characterize scRNA-seq data, and a soft self-training k-means algorithm is used to cluster cell populations in the potential learning space. The scBGEDA [13] performs single-cell clustering by a dual denoising autoencoder and bipartite graph ensemble clustering algorithm. Most of the models mentioned above use k-means clustering for initialization, and the results are optimized based on clustering loss.…”