Supplementary material: Prediction of protein-protein interactions using one-class classification methods and integrating diverse biological data

OCC methods can be classified according to the way in which they analyze, describe and gen-erate a model for the separation of targets and outlier examples [S1]. Two types of OCC were employed in this reseach: Density estimation methods based in the estimation of the probability density distribution of the training data using some probabilistic model (i.e. Gaussian distribu-tion), then a threshold is selected and then used to compare with the density of new objects in order to classify them; And Boundary methods, based in the generation of a frontier or bound-ary around the target objects, which is optimized to accept most of the target examples and at the same time reject most of the outliers. Boundary approaches are mainly focused on those examples objects which are located near the boundary. Here we present a detailed description of the four OCC approaches evaluated in this research: Gaussian density estimation: This is the simplest of the OCC density approaches. The examples of the target class used for training are modeled as a Gaussian distribution. In the dd tools implementation the complete density estimation is not obtained and just the Mahalanobis distance is employed and calculated for each example X as