Specter聚类方法在单细胞RNA测序数据分析中的应用  

The Application of Specter Clustering Method to Analysis in Single-cell RNA Sequencing Data

作　　者：何佳 王文杰[1] 张刘超 李康[1] He Jia;Wang Wenjie;Zhang Liuchao(Department of Medical Statistics,Harbin Medical University(150081),Harbin)

机构地区：[1]哈尔滨医科大学卫生统计学教研室,150081

出　　处：《中国卫生统计》2023年第3期354-357,362,共5页Chinese Journal of Health Statistics

基　　金：国家自然科学基金(81973149)。

摘　　要：目的研究Specter聚类的原理和方法,探讨其适用条件,并将其应用于卵巢癌单细胞测序(scRNA-seq)数据的聚类分析,标记得到的细胞亚群。方法通过模拟试验和实际单细胞RNA数据分析,对Specter聚类和较常用的其他三种方法进行比较,并用ARI和NMI两种评价指标的总得分进行评价。结果模拟试验显示,在设定的条件下,Specter聚类性能均优于其他三种方法;实例数据结果也表明,Specter聚类能够得到合理的细胞亚群。结论Specter聚类方法降低了参数敏感性,提高了聚类的准确性,具有研究价值和应用价值。Objective To research the principle and method of Specter clustering proposed recently,discuss its applicable conditions,and applies it to cluster analysis in ovarian cancer single cell sequencing(sRNA-seq)data,finally labels the cell subpopulations.Methods Compared Specter and the other three methods through simulation test and actual single-cell RNA data,then the total scores of ARI and NMI were used to evaluate the performance.Results As shown in Simulation tests,Specter clustering performance is better than the other three methods under the set conditions;The real data also showed that Specter could obtain reasonable cell subpopulations.Conclusion Specter reduces the sensitivity of parameters and improves the accuracy of clustering,which has research value and application value.

关 键 词：Specter scRNA-seq 聚类 细胞亚群 

分 类 号：R195.1[医药卫生—卫生统计学]