RANKCORR

Rankcorr: Rank-based marker selection for single-cell RNA sequencing

Rankcorr performs marker selection in high-throughput single-cell RNA sequencing (scRNA-seq) data by identifying small sets of genes that distinguish specific cell populations within datasets containing up to one million cells.

Key Features:

• Rank-Based Methodology: Applies rank correlation to mRNA counts data, using a non-parametric framework suited to count-based scRNA-seq measurements.
• Multi-Class Marker Selection: Selects markers across multiple cell classes using a structured, non-ad hoc approach.
• Computational Efficiency: Processes datasets ranging from thousands to one million cells with high speed and scalability.
• Performance Evaluation Metrics: Implements quantitative metrics to assess marker quality in the absence of known ground truth, enabling objective comparison with alternative marker selection methods.

Scientific Applications:

• Cell Type Identification: Identifies genetic markers for rare cell types and developmental pathways in large-scale scRNA-seq studies.
• Computational Pipelines: Integrates into scRNA-seq analysis workflows for large experimental and synthetic datasets.

Methodology:

Rankcorr ranks mRNA count data prior to analysis and performs linear separation of ranked expression profiles to select a minimal set of genes that optimally discriminates cell populations while maintaining interpretability.