Coralysis

Coralysis enables sensitive integration, reference-mapping, and cell-state identification in single-cell data via a multi-level integration algorithm.

Key Features:

• Multi-level integration algorithm: Progressively blends batch effects and separates cell types across multiple rounds of divisive clustering by grouping cells based on low-to high-level features.
• Imbalanced cell type handling: Outperforms state-of-the-art methods when similar cell types are unevenly distributed across batches or completely absent from some datasets.
• Cross-modality integration: Integrates rare cell populations from single-cell proteomic assays, such as basophils (0.5%) from whole blood.
• Reference-mapping and cell type prediction: Accurately predicts cell type identities across various query-reference scenarios, including reclassification of previously misclassified cell types such as CD16+ monocytes and natural killer cells.
• Cell-state probability scoring: Provides probability scores that enable identifying both transient and steady cell states along with their differential expression programs.

Scientific Applications:

• Single-cell RNA-seq integration: Demonstrates consistent performance across diverse single-cell RNA-seq integration tasks with heterogeneous datasets containing similar but unshared cell types.
• Single-cell proteomics integration: Enables integration of rare cell populations from single-cell proteomic assays.
• Cell state dynamics analysis: Facilitates study of subtle biological variation and its dynamics by improving integration of imbalanced cell types and states.
• Query-reference mapping: Reclassifies misclassified cell types in peripheral blood mononuclear cells, such as distinguishing CD16+ monocytes from CD14+ monocytes and natural killer cells from cytotoxic T cells.

Methodology:

The multi-level integration algorithm uses multiple rounds of divisive clustering inspired by puzzle assembly, beginning with grouping cells based on low-to high-level features before refining by shape and patterns.