Related Experiment Video
Updated: Jun 5, 2025
04:58
Author Spotlight: Investigating the Role of Repetitive DNA Misregulation in Cancer Initiation and Immunotherapy Resistance
Published on: December 13, 2024
2.1K
Analyzing scRNA-seq data by CCP-assisted UMAP and tSNE.
Yuta Hozumi1, Guo-Wei Wei1,2,3
1Department of Mathematics, Michigan State University, East Lansing, Michigan, United States of America.
Plos One
|December 13, 2024
Summary
Correlated clustering and projection (CCP) enhances single-cell RNA sequencing (scRNA-seq) data analysis by improving visualization and accuracy. This method preprocesses scRNA-seq data, boosting downstream machine learning tasks like UMAP and tSNE.
Area of Science:
- Genomics
- Bioinformatics
- Computational Biology
Background:
- Single-cell RNA sequencing (scRNA-seq) reveals cellular heterogeneity, crucial for understanding cell communication, differentiation, and gene expression.
- Analyzing scRNA-seq data presents challenges due to sparsity and high dimensionality, necessitating effective dimensionality reduction and feature selection.
- Correlated clustering and projection (CCP) is a novel preprocessing method for scRNA-seq data, leveraging gene-gene correlations and cell-cell interactions.
Purpose of the Study:
- To evaluate the efficacy of Correlated clustering and projection (CCP) as an initialization tool for dimensionality reduction techniques.
- To assess the impact of CCP on the visualization and accuracy of Uniform Manifold Approximation and Projection (UMAP) and t-distributed Stochastic Neighbor Embedding (tSNE).
- To demonstrate the generalizability of CCP across diverse scRNA-seq datasets.
Main Methods:
- Utilized Correlated clustering and projection (CCP) for preprocessing scRNA-seq data.
- Applied CCP as an initialization strategy for Uniform Manifold Approximation and Projection (UMAP) and t-distributed Stochastic Neighbor Embedding (tSNE).
- Validated the approach using 21 publicly available scRNA-seq datasets.
Main Results:
- CCP significantly enhanced the visualization quality of UMAP and tSNE.
- CCP dramatically improved the accuracy of UMAP and tSNE, with notable gains in Adjusted Rand Index (ARI), Normalized Mutual Information (NMI), and Efficiency-based Clustering Metric (ECM).
- Specifically, CCP improved UMAP by 22% in ARI, 14% in NMI, and 15% in ECM, and tSNE by 11% in ARI, 9% in NMI, and 8% in ECM.
Conclusions:
- Correlated clustering and projection (CCP) is an effective data-domain preprocessing method for scRNA-seq data.
- CCP initialization substantially improves the performance of UMAP and tSNE for scRNA-seq data analysis.
- The findings highlight CCP's potential to enhance downstream machine learning tasks in single-cell genomics.
