WEDGE

WEDGE (WEighted Decomposition of Gene Expression) performs biased low-rank matrix decomposition to impute sparse single-cell RNA sequencing (scRNA-seq) gene expression matrices and preserve biological correlations for downstream functional genomics analyses.

Key Features:

• Biased low-rank matrix decomposition: Implements a biased low-rank matrix completion algorithm that decomposes gene expression matrices into components capturing essential patterns while minimizing noise.
• Imputation of sparse matrices: Recovers missing values in highly sparse scRNA-seq expression matrices that exhibit high dropout rates.
• Preservation of biological correlations: Maintains both cell-wise and gene-wise correlations during imputation to retain inherent biological relationships.
• Improved clustering performance: Produces more complete expression representations that enhance clustering of cells for cell type and state identification.

Scientific Applications:

• Functional genomics analyses: Enables more comprehensive single-cell analyses of gene function and regulation by supplying imputed expression matrices.
• Cell type identification: Supports more precise identification and characterization of distinct cell populations via improved clustering.
• Disease research: Aids studies of heterogeneous diseases, including cancer and neurodegenerative disorders, by revealing subtle cellular differences.

Methodology:

Applies a biased low-rank matrix completion/decomposition algorithm to decompose scRNA-seq gene expression matrices into low-rank components and impute missing values while preserving cell-wise and gene-wise correlations.