DivBiclust

DivBiclust applies biclustering to identify cell subpopulations from single-cell RNA sequencing (scRNA-seq) transcriptomic profiles.

Key Features:

• Biclustering Approach: Simultaneously clusters genes and cells to identify subpopulations where specific sets of genes exhibit coherent expression patterns across subsets of cells.
• Handling High-Dimensional Data: Tailored to scRNA-seq datasets that measure gene expression across tens of thousands of genes per cell.
• Robustness to Noise and Sparsity: Accounts for technical noise and dropout events common in scRNA-seq to support reliable detection of cell subpopulations.
• Comparative Performance: Demonstrated superior accuracy in identifying cell subpopulations across ten real scRNA-seq datasets with varying sizes and dropout rates compared to nine state-of-the-art methods.
• Identification of Functionally Distinct Groups: Focuses on subsets of genes and cells to reveal functionally distinct cell groups that may be missed by global clustering methods.

Scientific Applications:

• Cellular Heterogeneity Analysis: Uncovers cell-specific transcriptomic changes and distinct cell types or states in heterogeneous samples.
• Developmental Biology: Supports analysis of cell type emergence and differentiation trajectories using scRNA-seq data.
• Cancer Genomics and Immunology: Identifies tumor cell subpopulations and immune cell states from transcriptomic profiles.

Methodology:

Implements a biclustering algorithm that optimizes gene and cell clustering simultaneously and focuses on subsets of genes and cells to capture complex expression patterns, with procedures described to account for noise and dropout in scRNA-seq data.