Figure 1.

Correlation between precision and the number of predicted residues (a) Training set; (b) Testing set. The prediction results of all the proteins in the datasets were calculated and averaged. The precisions of random prediction are 15% and 12.6% for the training and testing sets, respectively.