Unveiling gene modules at Atlas scale through hierarchical clustering of single-cell data

Area of Science:

• Computational Biology
• Genomics
• Machine Learning

Background:

• Single-cell RNA sequencing (scRNA-seq) analysis faces challenges in reconstructing cell hierarchies and identifying conserved gene modules across diverse datasets.
• Existing data integration and batch-correction methods struggle to resolve cell states in heterogeneous biological samples like tumors and immune tissues.

Purpose of the Study:

• To introduce Super Single-Cell Clustering (SuperSCC), a novel computational framework for discovering cell identities and gene modules across multiple scRNA-seq datasets.
• To enable the creation of hierarchical cell programs at both lineage and state levels, facilitating the identification of shared rare cell populations.
• To improve cell label mapping between reference and query datasets, overcoming batch effect limitations.

Main Methods:

• SuperSCC employs machine learning models to analyze scRNA-seq data, enabling cross-dataset discovery without requiring data integration.
• The framework supports analysis at both cell lineage and cell state levels, creating hierarchical representations of cellular programs.
• SuperSCC was applied to over 90 datasets for atlas-level analysis and mapping cell states in complex tissues.

Main Results:

• SuperSCC successfully identified cell identities and conserved gene modules across numerous scRNA-seq datasets, irrespective of batch effects.
• The framework demonstrated superior performance in identifying cellular contexts and achieved higher annotation accuracy compared to existing methods.
• Analysis of human lung tissue revealed conserved immune cell states in both healthy and diseased microenvironments.

Conclusions:

• SuperSCC provides a robust computational framework for advanced scRNA-seq data analysis, overcoming limitations of current integration methods.
• The ability to create hierarchical cell programs and identify shared rare populations offers significant advantages for understanding complex biological systems.
• SuperSCC enhances the accuracy of cell annotation and reveals conserved cellular states, particularly within the immune microenvironment of complex tissues.