Learning Consistency and Specificity of Cells From Single-Cell Multi-Omic Data.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell technologies generate epigenomic (DNA methylation, scATAC-seq) and transcriptomic (scRNA-seq) data.
• Integrating multi-omics single-cell data is challenging due to data heterogeneity, complex coupling, and interpretability issues.

Purpose of the Study:

• To propose a novel algorithm, self-representation Learning-based Multi-omics data Integrative Clustering (sLMIC), for effective integration and clustering of single-cell multi-omics data.
• To explicitly extract consistent and specific cell features for improved cell clustering.

Main Methods:

• sLMIC transforms omics data into multi-layer networks by constructing graphs for each data type, addressing heterogeneity.
• It utilizes low-rank and exclusivity constraints to separate cell self-representation into shared and specific features.
• Feature extraction and cell clustering are jointly optimized within a unified objective function.

Main Results:

• sLMIC successfully integrates single-cell epigenomic and transcriptomic data.
• The algorithm effectively models cell type structures by capturing both shared and specific features.
• Experiments on 13 diverse multi-omics datasets demonstrate sLMIC's superior performance over existing methods.

Conclusions:

• sLMIC provides an effective strategy for integrating and analyzing single-cell multi-omics data.
• The method enhances cell clustering by explicitly addressing data consistency and diversity.
• sLMIC offers a significant advancement in understanding biological mechanisms through integrated single-cell omics analysis.