Figure 3From: Metabolic network prediction through pairwise rational kernelsDiagram of pairwise SVM applied to metabolic network prediction. (a) An example of the pairs in the training set using the EC numbers (top) or gene names (bottom). (b) The pairwise kernel as a matrix, where the numerical values in each cell correspond to a measure of similarities, given two pairs of EC numbers (top) or two pairs of gene names (bottom). (c) A model is trained to estimate the parameters αi,j and b of the decision function f. (d) Given a new pair of EC numbers (left) or gene names (right) the decision function is evaluated and the pair is classified as interacting or non-interacting.Back to article page