Integrated analysis of multimodal single-cell data.

Area of Science:

• Single-cell genomics
• Immunology
• Computational biology

Background:

• Multimodal single-cell analysis is crucial for defining cellular states.
• Existing computational methods struggle with integrating diverse data types.
• There is a need for advanced frameworks to leverage multimodal datasets.

Purpose of the Study:

• To introduce a novel unsupervised framework, "weighted-nearest neighbor" (WNN) analysis.
• To enable integrative analysis of multiple modalities for single-cell data.
• To construct a multimodal reference atlas of the human circulating immune system.

Main Methods:

• Applied WNN analysis to a large CITE-seq dataset (211,000 human PBMCs).
• Utilized antibody panels for deep proteomic profiling alongside transcriptomic data.
• Developed a framework to learn the relative utility of each data type per cell.

Main Results:

• Substantially improved resolution of cell states through multimodal integration.
• Identified and validated novel lymphoid subpopulations.
• Constructed a comprehensive multimodal reference atlas of circulating immune cells.
• Demonstrated rapid mapping of new datasets and interpretation of immune responses.

Conclusions:

• WNN analysis offers a broadly applicable strategy for multimodal single-cell data.
• This approach moves towards a unified, multimodal definition of cellular identity.
• The developed atlas facilitates understanding of immune responses in vaccination and COVID-19.