Chongfan Technology
News
03
2024
-
09
The team of Wang Jianxin of Central South University proposed scCAD: Application of rare cell abnormality detection based on clustering decomposition in single cell expression data
Author:
Single-cell RNA sequencing (scRNA-seq) technology has become an important tool for characterizing cellular phenomena within complex tissues.Large-scale single-cell transcriptomics has great potential for identifying rare cell types that are critical to disease pathogenesis and biological processes. Existing methods for identifying rare cell types typically rely on one-time clustering using partial or global gene expression.However, these rare cell types may be overlooked at the clustering stage, which poses a challenge to their accurate identification.
On August 31, 2024,Central South UniversityWang Jianxin and Li Hongdong Joint CommunicationInNature Communications Published online entitled“scCAD: Cluster decomposition-based anomaly detection for rare cell identification in single-cell expression data”The research paper,The studyAn anomaly detection method based on cluster decomposition (scCAD) is proposed, which iteratively decomposes the clusters according to the most different signals in each cluster, effectively separates rare cell types, and realizes accurate recognition.
The study benchmarked scCAD on 25 real scRNA-seq datasets, demonstrating superior performance compared to 10 state-of-the-art methods. The in-depth case studies span diverse datasets, including mouse airway, brain, intestine, human pancreas, and immunological data, demonstrating the efficiency of scCAD in identifying rare cell types in complex biological scenarios.In addition, scCAD can correct annotations of rare cell types and identify disease-associated immune cell subtypes, providing valuable insights into disease progression.
Single-cell RNA sequencing (scRNA-seq) technology enables researchers to analyze gene expression patterns at the single-cell level, dissecting cellular heterogeneity while providing new insights into the composition and function of cell types in complex tissues.As sequencing technology advances, larger datasets become available that can not only characterize major cell types, but also capture low-frequency cell types. These rare cell types exhibit low abundance and have been widely demonstrated to play important roles in disease pathogenesis and biological processes, such as angiogenesis and immune response mediation. For example, circulating tumor cells (CTCs) are indeed very rare in peripheral blood, but their metastasis is strongly associated with cancer-related deaths. It is estimated that every 105-106Peripheral blood mononuclear cells (PBMCs) have 1 or less CTCs cells. This limitedCTCsPresence presents significant challenges for its detection and characterization in cancer research. Therefore,Developing specialized methods to accurately and efficiently identify and characterize these rare cell types has become a major challenge for single-cell research.
In recent years, the main algorithms for the identification and analysis of rare cell types include FiRE, CellSIUS, EDGE for simultaneous dimension reduction and feature gene extraction, GapClust, GiniClust series methods, RaceID series methods, SCISSORS,CIARAand SCA.These methods mainly identify rare cells from four aspects. The first idea involves proposing a method for measuring the rarity of cells in a highly variable gene space. The second perspective focuses on proposing a feature selection process. The CIARA algorithm uses KNN to identify potentially rare cells by detecting highly locally expressed genes, and then applies the Louvain algorithm to cluster with the selected genes. The third perspective is based on the clustering results, and a customized method for rare sub-clustering is proposed. CellSIUS identify candidate marker genes with expression values in each clusterbimodal distributionand then perform one-dimensional k-means clustering based on the average expression of each gene set with associated expression patterns, further dividing the cells into sub-clusters. RaceID identifies abnormal cells within a cluster by assessing transcript count variability for each gene across all cells, and then reassigns each cell into the cluster with the highest correlation. The last point of view involves proposing dimensionality reduction methods for distinguishing rare cells, such as EDGE and SCA.Furthermore, the integration of multi-omics data has emerged as a promising approach. For example, MarsGT combines scRNA-seq and scATAC-seq data for rare cell identification using a probabilistic allometric transducer.
Although some success has been achieved, these algorithms have limitations in terms of accuracy and robustness.Methods that rely on highly variable genes may be sensitive to the number of differentially expressed genes by ignoring specific signals that are important for distinguishing rare cell types. In identifying rare cell subpopulations, feature selection-based approaches ignore the potential dependencies between different genes. Cluster-based methods may require further analysis to distinguish between rare types of genes in each cluster. The dimensionality reduction method may lose important information in the process or be susceptible to noise and interference, which complicates the accurate identification of rare cells.Methods of integrating multi-omics data require consideration of potential noise, which can complicate the identification of rare cell types.
To overcome these limitations, scCAD, an anomaly detection method based on cluster decomposition, is proposed to effectively identify rare cell types.Compared with existing algorithms, scCAD provides an integrated feature selection method designed to maximize the preservation of differential signals of rare cell types. In the process of cluster decomposition, scCAD performs iterative clustering according to the most different signals in each cluster, effectively distinguishing rare types or subtypes that are initially difficult to distinguish. Finally, scCAD provides the user with several clusters of cells that may be rare.
The study benchmarked scCAD on 25 real-world scRNA-seq datasets, demonstrating its remarkable ability to identify rare cell types.In most of these data sets, scCAD shows higher recognition accuracy than other state-of-the-art methods. In mouse airway, brain, intestine, human peripheral blood mononuclear cells (PBMC) And in case studies of multiple biological scenarios including the pancreas, scCAD accurately identified the rare cell types reported in previous studies, showing its robustness and accuracy. Inclear cell renal cell carcinomascCAD corrects rare cell annotation errors and identifies disease-relevant immune cell subtypes, providing valuable insights into disease progression.In addition, the analysis of the results of two large immunological data sets shows that scCAD has good scalability.
Chongqing Chongfan Technology Co., Ltd. Optical platform film thickness meter OCT optical coherence tomography LabVIEW control
LATEST NEWS
Thank you for visiting the official website of Chongfan Technology. If you have cooperation intentions or suggestions, please contact us through the following methods, and we will reply as soon as possible, thank you!
Address: Room 403, Building 6, Phase III of R&D, No. 36 Xiyong Avenue, High tech Zone, Chongqing, China.
Telephone: +86-13658337211
E-mail: Sales@cfkeji.net
Website: www.cfkeji.net
Mobile Version
