scds

scds is an R package that provides two new computational methods, cxds and bcds, for identifying doublets (two or more cells wrongly considered single cells) in single-cell RNA sequencing (scRNA-seq) data. Doublets can bias downstream analyses and lead to incorrect conclusions, making their identification crucial.

The co-expression-based doublet scoring (cxds) method uses binarized gene expression data and a binomial model to assess the co-expression of gene pairs, providing interpretable doublet annotations. The binary classification-based doublet scoring (bcds) method uses a binary classification approach to distinguish artificial doublets from original data.

When applied to four datasets with ex

perimental doublet annotations and compared to existing doublet identification methods, scds performs at least as well as the state-of-the-art approaches while being computationally efficient. The package can annotate doublets in datasets with thousands of cells within seconds.