【摘要】 随着单细胞转录组测序技术的发展，无标签数据集也在日益俱增，然而细胞标签标注是一项耗时耗力的工作。提出了一种基于半监督学习的自编码器和图卷积神经网络单细胞分类算法，称为sctAGCN(single cell transcriptomics data classification via autoencoder and graph convolutional network)。首先，使用自编码器(autoencoder, AU)克服数据的高维性难题，将高维数据投射到低维空间。其次，借助相互最近邻算法(mutual nearest neighbor algorithm, MNN)寻求每个细胞的k个最近邻用于构造邻接矩阵。最后，将图卷积神经网络(graph convolutional neural network, GCN)作为分类器，用于单细胞分类。本工作使用跨测序方式和跨物种收集的5个数据集对模型性能进行了评估，结果表明sctAGCN能够有效提取单细胞信息，并且在实验中优于其它单细胞分类方法。更多还原

【Abstract】 With the development of single-cell transcriptomics sequencing technology, unlabeled data sets are also increasing. However, cell labeling is a time-consuming and labor-intensive task. This paper proposes a single cell classification algorithm based on semi-supervised learning autoencoder and graph convolutional neural network called sctAGCN(Single cell transcriptomics data classification via autoencoder and graph convolutional network). First, autoencoder(AU) is used to overcome the high-dimensionality problem of data and project high-dimensional data into low-dimensional space. Secondly, the k nearest neighbors of each cell is searched by mutual nearest neighbor algorithm(MNN) to construct the adjacency matrix. Finally, the graph convolutional neural network(GCN) is used as a classifier for single cell classification. In this paper, the performance of the model is evaluated based on five datasets collected by cross-sequencing and cross-species. The results show that sctAGCN can effectively extract single cell information and is superior to other single cell classification methods in experiments.更多还原