Fig. 13From: Analysis of the benefits of imputation models over traditional QSAR models for toxicity predictionAuxiliary assay selection for the assay ‘TOX21-Aromatase-Inhibition’. Auxiliary assays were selected according to the MI-entropy ratio (‘similarity_added’) or randomly (‘random_added’). As a comparison, performance of XGB as a single task model and the respective multi-task model using all remaining auxiliary assays were included. Plotted are the MCC scores for the test set across 20 runs with different random seeds. A XGB-FN models. B Macau modelsBack to article page