Identifying similar populations across independent single cell studies without data integration

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Single-cell data analysis presents challenges due to large, independent study pools.
• Existing methods often require data correction or integration, potentially introducing artifacts.

Purpose of the Study:

• To introduce ClusterFoldSimilarity (CFS), a novel statistical method for quantifying cell group similarity across multiple independent datasets.
• To enable cross-dataset comparison without data integration or correction, preserving information and avoiding artifacts.

Main Methods:

• CFS quantifies similarity between cell groups across datasets.
• It identifies conserved phenotypes and performs feature selection for cross-dataset markers.
• The method supports multimodal data, including single-cell RNA-Seq, ATAC-Seq, and proteomics.

Main Results:

• CFS successfully identified conserved astrocyte subpopulations in mouse motor cortex and spinal cord single-nuclei RNA-Seq data.
• The method demonstrated its ability to match cell groups with conserved phenotypes across different tissues and species.
• Feature selection identified cross-dataset markers for similar cell phenotypes, enhancing interpretability.

Conclusions:

• CFS offers a simple, efficient, and scalable solution for analyzing complex single-cell data from independent studies.
• The method facilitates the discovery of conserved cell populations and their defining features across diverse datasets.
• CFS provides visualization tools for interpreting similarity scores and cell population relationships.