Evaluation of our method with respect to internal and external prediction of the dataset.

(A) Effect of rational negative design. (B) Effect of the second-layer SVM with designed negatives. (C) Improvement of precision with the two-layer SVM ant the type of the first-layer SVM models.
†“Model type” exhibits the one-layer SVM model or the second-layer SVM, which is specified by the type of 11 first-layer SVM model, was utilized. Here,
• (designed) means that the rationally designed negatives was used to construct the SVM model.
• (random) means that three types of randomly chosen 22,050 pairs of protein and chemical compounds were used use to construct the SVM model. The 95% confidence intervals were shown.
• (r.f.) means that twenty types of randomly chosen 11 first-layer SVM models were used to construct the second-layer SVM model.
• 2nd ANN means that Artificial Neural Network (ANN; implemented by the statistical software package R (http://cran.r-project.org/) function nnet[15]) was applied to outputs of 11 subpos first-layer SVM models. Parameters were selected to give the best accuracy in internal 10-fold cross validation. For example, 17 units were used in the hidden layer.
• voting means that voting with 11 subpos first-layer SVM models was used for prediction.
• 2nd QDA means that Quadratic Discriminant Analysis (QDA) (implemented by R function qda[15]) was applied to outputs of 11 subpos first-layer SVM models.
• (t = 0.9) means that final probability outputs were evaluated with the threshold t = 0.9.
‡precision (prec.) = TP/(TP+FP), sensitivity (sens.) = TP/(TP+FN), accuracy (acc.) = (TP+TN)/(TP+FN+TN+FP). TN: true negatives. Here,
• ex means the prediction performances of the external prediction. The external dataset consisted of 170 positives and 2,450 negatives that were randomly chosen from 1,731 positives and 24,500 designed negatives with the mlt rule (details are provided in Materials and Methods) and that were excluded in constructing first-layer and second-layer SVM models.
• in means the prediction performances of internal 10-fold cross-validation. The internal dataset utilized 1,561 positives and 22,050 negatives, which were not included in the external dataset.