Removes doublets from single-cell RNA-seq data. DoubletDecon includes specialized methods to “rescue” preliminarily “removed” cells and cell-state clusters that include unique gene expression patterns. This approach is applicable to large and small datasets with both discrete cell populations or gradual cellular transitions, by automatically grouping correlated cells states without merging.
Detects doublets in single-cell RNA-seq (scRNA-seq) count matrices. DoubletDetection integrates artificial doublets into real data and computes the proportion of artificial nearest neighbors (pANN) for every real cell. It also identifies false-negative and putative false-positive doublet classifications. This method can be applied in experimental contexts where doublet formation rates differ significantly from industry estimates such as clumpy single-cell suspensions or especially cohesive cell types.
Discovers doublets in single-cell RNA sequencing data. DoubletFinder contains four main features: it (1) produces artificial doublets from existing scRNA-seq data; (2) performs principal component analysis (PCA) on the merged real-artificial; (3) determines the nearest neighbors for every real cell in PC and finds each cell’s proportion of artificial nearest neighbors (pANN) and ultimately (4) ranks order and thresholds pANN values according to the expected number of doublets.
Predicts impact of multiplets in a given analysis and identifying problematic multiplets. Scrublet is a method that includes some measure of ground truth labels for cell multiplets. It works in two steps: (i) multiplets are simulated from the data by combining random pairs of observed transcriptomes, and (ii) each observed transcriptome is scored based on the relative densities of simulated doublets and observed transcriptomes in its vicinity.