BRAQUE: Bayesian Reduction for Amplified Quantization in UMAP Embedding.

Area of Science:

• Single-cell biology
• Immunofluorescence imaging
• Spatial transcriptomics

Background:

• Single-cell biology has transformed biological understanding.
• Analyzing spatial single-cell data from immunofluorescence imaging requires tailored approaches.
• Existing methods may lack the granularity needed for complex biological systems.

Purpose of the Study:

• To present Bayesian Reduction for Amplified Quantization in UMAP Embedding (BRAQUE), an integrative approach for spatial single-cell data analysis.
• To improve data preprocessing and clustering for enhanced cellular phenotype classification.
• To achieve higher granularity in cell type identification compared to existing algorithms.

Main Methods:

• BRAQUE employs Lognormal Shrinkage for data preprocessing, fitting a lognormal mixture model to enhance fragmentation.
• Dimensionality reduction is performed using Uniform Manifold Approximation and Projection (UMAP).
• Clustering is achieved using Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) on UMAP embeddings, followed by expert-guided cell type assignment using effect size measures.

Main Results:

• Lognormal Shrinkage improves cluster separation and clarity.
• BRAQUE achieves higher granularity in cell type identification than PhenoGraph.
• The method facilitates the identification of Tier 1 and Tier 2 characterizing markers for cell types.

Conclusions:

• BRAQUE offers a robust and refined pipeline for spatial single-cell data analysis.
• The approach enables a more detailed understanding of cellular heterogeneity in tissues.
• BRAQUE's ability to reveal finer cellular distinctions advances single-cell biology research.