The EdgePrediction library is a Python machine learning library for knowledge graph completion, described and applied to the prediction of adverse drug reactions (ADRs) in Bean et al. 2017. The algorithm is based on using Fisher’s exact test for enrichment to identify properties of positive examples in the training data that are likely to be predictive. Using these features, a grid search of parameters determines feature weights that maximise the objective function (Louden’s J statistic). This trained model predicts edges that are missing from the graph.