Skip to main content
. 2022 May 30;5:510. doi: 10.1038/s42003-022-03476-9

Table 1.

Comprehensive performance of each method in real-world scRNA-seq datasets with experimentally annotated doublets.

PAUC800 PAUC900 PAUC950 PAUC975 AUC PR
bcds 0.598456581 0.697609168 0.747471204 0.772440146 0.797428571 0.465471429
Chord 0.602189241 0.701382997 0.751248502 0.77621856 0.801214286 0.464642857
ChordP 0.614164051 0.713609623 0.763485434 0.788453732 0.8132 0.466514286
cxds 0.576279854 0.675031498 0.724828357 0.749788808 0.774785714 0.367342857
doubletCells 0.396983835 0.487445535 0.535271112 0.559821949 0.584685714 0.173985714
DoubletDetection 0.569370526 0.666356657 0.715610186 0.740426505 0.793114286 0.500857143
DoubletFinder 0.537830717 0.636466839 0.686221775 0.711172069 0.736171429 0.339428571
Scrublet 0.564203075 0.663581473 0.713449225 0.738419459 0.763414286 0.399771429
Solo 0.604236942 0.703224153 0.752987364 0.777930394 0.803142857 0.434714286

The average performance of various methods in all datasets. The indexes are the pAUC800, pAUC900, pAUC950, pAUC975, AUC and AUPRC.