PsiNorm

PsiNorm normalizes single-cell RNA-seq (scRNA-seq) count data between samples by estimating a power-law Pareto distribution parameter to produce comparable inputs for downstream analyses.

Key Features:

• Normalization Methodology: Applies a between-sample normalization based on a power-law Pareto distribution parameter estimate, leveraging that scRNA-seq data can be well represented by the Pareto distribution regardless of sequencing depth or technology.
• Scalability and Efficiency: Implements an approach designed to handle large scRNA-seq datasets while addressing memory usage and computational time constraints.
• Benchmarking Performance: In comparative analyses against seven other normalization methods, demonstrated superior cluster identification accuracy and concordance with fewer computational resources.
• Reference-Free Normalization: Performs normalization without requiring a reference dataset, enabling normalization of out-of-sample data in supervised classification settings.

Scientific Applications:

• Tissue composition: Normalizes scRNA-seq data to support analyses of cell-type proportions and tissue composition.
• Cellular dynamics: Enables comparative analyses of cellular state changes and dynamics across samples.
• Organism development: Supports normalization for studies of developmental processes using scRNA-seq data.
• Cross-technology and depth integration: Facilitates normalization across different sequencing depths and sequencing technologies to improve comparability and biological interpretation.

Methodology:

PsiNorm estimates the Pareto (power-law) distribution parameter per sample and applies a between-sample normalization using that parameter.