Table 1.

Performance on benchmark datasets for on-target cleavage efficiency prediction

Dataset	Model	Metric	Value
DeepHF wildtype	(27), RNN	Hold-out SCC	0.8555
DeepHF wildtype	(27), RNN	10-fold CV SCC	NA
DeepHF wildtype	This study, C E	Hold-out SCC	0.8392
DeepHF wildtype	This study, C E	10-fold CV SCC	0.8066
DeepHF eSpCas9	(27), RNN	Hold-out SCC	0.8491
DeepHF eSpCas9	(27), RNN	10-fold CV SCC	NA
DeepHF eSpCas9	This study, R E	Hold-out SCC	0.8220
DeepHF eSpCas9	This study, R E	10-fold CV SCC	0.6927
DeepHF SpCas9-HF1	(27), RNN	Hold-out SCC	0.8512
DeepHF SpCas9-HF1	(27), RNN	10-fold CV SCC	NA
DeepHF SpCas9-HF1	This study, R E+M	Hold-out SCC	0.8364
DeepHF SpCas9-HF1	This study, R E+M	10-fold CV SCC	0.7900
geCRISPR V520	(26), mono binary	Hold-out PCC	0.6700
geCRISPR V520	(26), mono binary	10-fold CV PCC	0.6800
geCRISPR V520	This study, C E+M	Hold-out PCC	0.6055
geCRISPR T3619	This study, C E+M	10-fold CV PCC	0.5926
DeepCpf1 H1	(23)	Hold-out SCC	0.7600
DeepCpf1 H1	This study, R E	Hold-out SCC	0.7283
DeepCpf1 H2	(23)	Hold-out SCC	0.7400
DeepCpf1 H2	This study, C E+M	Hold-out SCC	0.7184
DeepCpf1 H3	(23)	Hold-out SCC	0.5800
DeepCpf1 H3	This study, R E	Hold-out SCC	0.5478
DeepCpf1 train	(23)	10-fold CV SCC	NA
DeepCpf1 train	This study, C E+M	10-fold CV SCC	0.5165