Table 8.

Comparing unified RNN-CNN and unified RNN/GCNN-CNN based on RMSE (and Pearson’s r) for pIC₅₀ prediction

	SMILES rep.			Graph rep.
	Single	Parameter	Parameter+NN	Single	Parameter	Parameter+NN
		ensemble	ensemble		ensemble	ensemble
Training	0.47 (0.94)	0.45 (0.95)	0.44 (0.95)	0.55 (0.92)	0.54 (0.92)	0.55 (0.92)
Testing	0.78 (0.84)	0.77 (0.84)	0.73 (0.86)	1.50 (0.35)	1.50 (0.35)	1.34 (0.45)
Generalization—ER	1.53 (0.16)	1.52 (0.19)	1.46 (0.30)	1.68 (0.05)	1.67 (0.03)	1.67 (0.07)
Generalization—Ion Channel	1.34 (0.17)	1.33 (0.18)	1.30 (0.18)	1.43 (0.10)	1.41 (0.13)	1.35 (0.12)
Generalization—GPCR	1.40 (0.24)	1.40 (0.24)	1.36 (0.30)	1.63 (0.04)	1.61 (0.04)	1.49 (0.07)
Generalization—Tyrosine Kinase	1.24 (0.39)	1.25 (0.38)	1.23 (0.42)	1.74 (0.01)	1.71 (0.03)	1.70 (0.03)

Note: Bold-faced entries correspond to the best performance for each row (data set).