PanoView

PanoView performs iterative clustering of single-cell RNA sequencing (scRNA-seq) data to identify and characterize major and rare cell subpopulations.

Key Features:

• Iterative PCA-based clustering: Uses principal component analysis to reduce dimensionality and iteratively refines clusters by selecting confident cell groupings and recalculating PCA for remaining cells.
• Density-based OLMC algorithm: Applies Ordering Local Maximum by Convex Hull (OLMC) to detect local density maxima and order clusters based on convex hull relationships to handle varying cell densities.
• Heuristic parameter estimation: Adapts parameter values from input data structures via a heuristic estimation approach.
• Simultaneous detection of major and rare cell types: Identifies both prevalent and rare cell populations within the same analysis.

Scientific Applications:

• Validation on simulated and published scRNA-seq datasets: Tested on simulated datasets and published single-cell RNA-sequencing datasets with reported improvements in accuracy and reliability compared to other methods.
• Embryonic mouse hypothalamus analysis: Applied to embryonic mouse hypothalamus scRNA-seq data to recover known cell types and detect rare subpopulations.

Methodology:

PanoView applies PCA to transform high-dimensional scRNA-seq data, uses the OLMC algorithm to identify and order local density maxima based on convex hulls, and iterates by extracting confident clusters and recalculating PCA on remaining cells until subpopulations are delineated.