Doublets (i.e. two cells captured as a single cell) can form at a high frequency in single-cell sequencing studies (often 10-20%). They can appear as spurious ‘new cell types’ or distort trajectory and co-expression analyses. While experimental strategies can be used to mitigate this effect, they are insufficient and need to be complemented with doublet identification methods. In this talk I discuss the main doublet identification strategies, and present the scDblFinder package and its main method. I demonstrate its superiority to alternatives in terms of speed and accuracy on a vast majority of datasets, and then turn to the problem of doublet characterization and doublet enrichment analysis.