基于对比学习的单细胞转录组测序数据聚类模型  

A Clustering Model For ScRNA-seq Data Based on Contrastive Learning

作　　者：张珊珊 林劼[1] ZHANG Shanshan;LIN Jie(School of Mathematics and Statistics,Fujian Normal University,Fuzhou 350117,China)

机构地区：[1]福建师范大学数学与统计学院,福建福州350117

出　　处：《福建师范大学学报（自然科学版）》2023年第2期26-34,共9页Journal of Fujian Normal University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61472082)。

摘　　要：单细胞转录组测序技术(single-cell RNA sequencing,scRNA-seq)的快速发展为分析生物数据提供了有力支持.对scRNA-seq数据进行聚类分析,能够发现潜在的细胞亚型并研究细胞的异质性.但由于scRNA-seq数据存在高维性、高稀疏性以及dropout事件等问题,为聚类分析带来了挑战.提出一种基于对比学习的聚类方法,假设数据服从零膨胀负二项分布,应用自编码器框架学习细胞的表示.实验结果表明提出的方法在真实数据集上有优越的性能,在不同规模的数据集上具有良好的可扩展性.The rapid development of single-cell RNA sequencing(scRNA-seq)technology provides strong support for the analysis of biological data.Cell clustering in scRNA-seq analysis can identify potential cell subtypes and study cell heterogeneity.However,high dimensionality,high sparsity and dropout events of scRNA-seq data produces challenges in clustering analysis.This paper proposes a clustering method based on contrastive learning,which assumes that the data follows a zero inflated negative binomial distribution and uses an autoencoder framework to learn the representation of cells.Experimental results show that the method has excellent performance on real datasets and good scalability on different scale datasets.

关 键 词：单细胞转录组测序数据 对比学习 零膨胀负二项分布模型 自编码器 

分 类 号：Q811.4[生物学—生物工程]