Detect doublets in single-cell RNA-seq data by simulating doublets from the observed transcriptomes and scoring each cell with a k-nearest-neighbor classifier.