RgCop

RgCop selects predictive, non-redundant gene features from single-cell RNA sequencing (scRNA-seq) datasets using copula-based dependence measures to support clustering, classification, and cell annotation.

Key Features:

• Regularized Copula Correlation (Ccor): RgCop employs copula correlation (Ccor) to capture multivariate dependencies among genes in single-cell RNA sequencing (scRNA-seq) data.
• l1 Regularization: An l1 regularization term penalizes redundant feature coefficients to yield a non-redundant and stable set of genes.
• Scale-invariant Copula Property: The scale invariant property of copula functions is leveraged to maintain robustness to technical noise in scRNA-seq measurements.

Scientific Applications:

• Clustering and Classification: Selected genes improve clustering and classification performance on scRNA-seq datasets.
• Differential Expression and Cell Annotation: Identified differentially expressed (DE) genes support accurate cell annotation from scRNA-seq data.
• Cellular Heterogeneity Analysis: Robust, non-redundant feature sets enable discovery of cellular heterogeneity and interpretation of complex biological processes in single-cell studies.

Methodology:

RgCop computes copula-based correlation (Ccor) to assess multivariate gene dependencies and applies an l1 regularization term to penalize redundant feature coefficients, leveraging the scale-invariant property of copula functions.