CellBouncer, a unified toolkit for single-cell demultiplexing and ambient RNA analysis, reveals hominid mitochondrial incompatibilities

Pooled processing, in which cells from multiple sources are cultured or captured together, is increasingly popular for droplet-based single-cell sequencing studies. This design allows efficient scaling of experiments, isolation of cell-intrinsic differences, and mitigation of batch effects. We present CellBouncer, a computational toolkit for demultiplexing and analyzing single-cell sequencing data from pooled experiments. We demonstrate that CellBouncer can separate and quantify multi-species and multi-individual cell mixtures, identify unknown mitochondrial haplotypes in cells, assign treatments from lipid-conjugated barcodes or CRISPR single-guide RNAs, and infer pool composition, outperforming existing methods. We introduce methods to quantify ambient RNA contamination per cell, infer individual donors' contributions to the ambient RNA pool, and determine a consensus doublet rate harmonized across data types. Applying these tools to tetraploid composite cells, we identify a competitive advantage of human over chimpanzee mitochondria across ten cell fusion lines and provide evidence for inter-mitochondrial incompatibility and mito-nuclear incompatibility between species.