scds

scds computationally annotates doublets in single cell RNA sequencing (scRNA-seq) data using complementary co-expression based and binary classification based scoring approaches.

Key Features:

• cxds (Co-expression based doublet scoring): Utilizes binarized gene expression data and employs a binomial model for the co-expression of pairs of genes to yield interpretable doublet annotations.
• bcds (Binary classification based doublet scoring): Uses a binary classification approach to discriminate artificial doublets from the original data.
• Scalability: Enables doublet annotations in thousands of cells with comparably little computational cost.

Scientific Applications:

• Doublet detection in high-throughput scRNA-seq: Identifies doublets in experiments employing micro-fluidics in combination with unique molecular identifiers (UMIs) that assay many thousands of cells.
• Bias reduction in downstream analyses: Prevents severe bias in study conclusions that results from treating doublets as single cells in downstream analyses.

Methodology:

The cxds method binarizes gene expression data to absence/presence calls and applies a binomial model for co-expression of gene pairs. The bcds method generates artificial doublets and uses binary classification to discriminate them from original data.