Unsupervised topological alignment for single-cell multi-omics integration.

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Single-cell multi-omics data offer deep molecular insights but present integration challenges due to unpaired cells and unmatched features across modalities.
• Current methods struggle with the inherent complexity and lack of direct correspondence in multi-omics datasets.

Purpose of the Study:

• To develop a novel algorithm, UnionCom, for unsupervised topological alignment of single-cell multi-omics data.
• To enable robust integration of datasets with distinct features and dataset-specific cell types without requiring cell or feature correspondence.

Main Methods:

• UnionCom embeds intrinsic low-dimensional structures into cell distance matrices for each dataset.
• It aligns cells across datasets by matching distance matrices using matrix optimization and a global scaling parameter.
• Distinct features are projected into a common embedding space for comparability.

Main Results:

• UnionCom successfully integrates single-cell multi-omics data by aligning topological structures.
• The algorithm outperforms existing state-of-the-art methods on both simulated and real datasets.
• UnionCom demonstrates robustness to parameter choices and feature subsampling.

Conclusions:

• UnionCom provides an effective unsupervised approach for single-cell multi-omics integration.
• The method's ability to handle complex data structures and dataset-specific cell types advances the field.
• The software is publicly available for broader research application.