SCA

SCA applies surprisal-based dimensionality reduction to single-cell transcriptomic datasets to detect rare and subtle cellular subpopulations and to improve downstream imputation.

Key Features:

• Surprisal-based dimensionality reduction: Uses the information-theoretic concept of surprisal to derive components from single-cell transcriptomic data.
• Detection of subtle variations: Emphasizes small or nuanced transcriptional differences that are often missed by conventional dimensionality reduction methods.
• Identification of rare subpopulations: Highlights rare or clinically significant cell populations, including specific cytotoxic T-cell subsets.
• Improved imputation: Produces representations that improve downstream imputation and reconstruction of missing transcriptomic values.
• Information-theoretic framework: Employs an efficient information-theoretic approach for signal extraction from complex biological datasets.

Scientific Applications:

• Single-cell transcriptomic analysis: Characterizing cellular heterogeneity within single-cell RNA-seq datasets.
• Detection of rare cell types: Identifying rare or small cellular subpopulations such as cytotoxic T-cell subsets.
• Study of complex tissues: Analyzing cellular composition and variation in complex biological tissues in health and disease contexts.
• Support for therapeutic research: Providing enhanced signals that inform personalized medicine and therapeutic intervention studies.

Methodology:

Computes surprisal scores from single-cell transcriptomic measurements and applies surprisal-based dimensionality reduction, with resulting components used to improve downstream imputation.