Figure 3.

Inconsistent predictions for the s domain of casein kinase 1. The graph shows a portion of the molecular function Gene Ontology (GO), corresponding to positive labels and labels predicted with probability larger than 0.1. Each term's outline is colored according to the label with respect to the target term 'protein-tyrosine kinase activity'. A green outline corresponds to a positive label, and a purple outline corresponds to a negative label. For a given protein from the held-out data set, the interior of each node is colored following a similar color scheme (green = 1 and purple = 0), according to the probability produced by several data specific support vector machines (SVMs) combined into a logistic regression.