Table 2.

Performance comparison of CNN-SVR and different CNNs combined regression models for gRNA activity prediction on four cell-line datasets under 10-time 10-fold cross-validation.

Model	HCT116	HEK293T	HELA	HL60
(A) Spearman correlation
CNN-SVR	0.719 ± 0.008	0.807 ± 0.016	0.699± 0.006	0.589± 0.006
CNN-L1	0.712± 0.010	0.793 ± 0.004	0.633± 0.020	0.542± 0.033
CNN-L2	0.670± 0.025	0.731 ± 0.032	0.683± 0.009	0.517± 0.034
CNN-L1L2	0.701± 0.008	0.803 ± 0.012	0.682± 0.005	0.589± 0.018
(B) AUROC
CNN-SVR	0.933 ± 0.001	0.983 ± 0.002	0.933 ± 0.001	0.934 ± 0.003
CNN-L1	0.931 ± 0.001	0.982 ± 0.001	0.924 ± 0.002	0.930 ± 0.003
CNN-L2	0.919± 0.002	0.975 ± 0.002	0.923± 0.002	0.895± 0.008
CNN-L2	0.918± 0.003	0.977 ± 0.001	0.915± 0.002	0.912± 0.004

The tables from top to bottom respectively record the Spearman correlation and AUROC of CNN-SVR and three CNN combined regression methods.