Distribution of predicted distance to the hyperplane. The distance to the hyperplane of a support vector machines model is related to the confidence of the predicted results. The bigger the distance is, the more confident the predicted results will be. The histogram is the prediction results for the drug targets and non drug targets in the 10-fold cross-validation of training set (4) – (positive/negative = 1:6). Drug target families contain 3,444 human original proteins from Swiss-Prot, which are in the same family of known drug targets. Non drug-target families consist of 9,758 putative non drug targets. Research drug targets contain 371 human origin research drug target proteins which do not belong to the known drug target family.