Single-Cell Multi-omics Clustering Algorithm Based on Adaptive Weighted Hyper-laplacian Regularization

Wei Lan, Shengzu Huang, Xun Sun, Haibo Liao, Qingfeng Chen, Junyue Cao

Abstract: Modern single-cell sequencing technologies are capable of analyzing multiple molecular patterns from the same single cell, which provides an unprecedented opportunity to analyze cellular heterogeneity from multiple biological levels. Clustering single-cell multi-omics data can provide deeper insights into cellular states and their regulatory mechanisms. However, existing single-cell clustering methods focus on single omics data and ignore higher-order information between different samples. In this paper, we proposed a new multi-view subspace single-cell clustering algorithm (scAHVC) for joint clustering analysis of single-cell ATAC-seq data and single-cell RNA-seq data. It performs low-rank representations of single-cell omics data by using tensor nuclear norm to obtain consistent information across omics. Then, the adaptive weighted hyper-laplacian regularization is used to preserve the local structure of the data in the high-dimensional space and fully explore the higher-order information of the data. The experimental results show that scAHVC outperforms the other state-of-the-art methods on clustering performance.