SMILE: mutual information learning for integration of single-cell omics data.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Deep learning enhances single-cell omics analysis, revealing complex cellular system insights.
• Integrating multisource, multimodal, and multi-feature single-cell data presents significant challenges.

Purpose of the Study:

• To introduce an unsupervised deep learning algorithm, Single-cell Mutual Information Learning (SMILE), for integrating diverse single-cell omics data.
• To develop a method for learning discriminative representations by maximizing mutual information.

Main Methods:

• SMILE employs an unsupervised deep learning approach.
• It utilizes a unique cell-pairing design to integrate data.
• Maximizing mutual information is key to learning representations.

Main Results:

• SMILE successfully integrates multisource single-cell transcriptome data, removing batch effects.
• It projects similar cell types from different tissues into a shared space.
• SMILE integrates multi-modal data (ATAC-seq, RNA-seq, DNA methylation, Hi-C, ChIP) and data with unmatched features.

Conclusions:

• SMILE provides a robust framework for integrating diverse single-cell omics data.
• The learned representations facilitate comparisons of independent single-source data.
• The algorithm is available in Python with analysis code and supplementary data.