Table 1.

Class-specific network and PSSM results. The comparison shows the substantial improvement of the machine learning method with respect to PSSM. The higher sensitivity and precision of the neural model with respect to the PSSM indicate that the former is able to predict a lower number of false positives. The higher specificity of the neural model also implies a better filtering of non-interacting sequences and a higher performance of the model in the detection of SH3 binders.

Class	Number of Binders	PSSM				NN
		Prec	Sens	Spec	Corr	Prec	Sens	Spec	Corr
I	88 (13.1%)	40	73	84	0.45	51	77	89	0.56
II	131 (18.5%)	52	64	87	0.47	57	72	88	0.55