Fig. 2

The box plotting of validation performance for the single task models and multitask models. STL-AG and STL-ATG presented the performances of single-task models trained using the individual receptor with agonist or antagonist datasets; MTL-AG and MTL-ATG for the performances of the multitask models trained using integrated receptors with agonist or antagonist datasets, the MTL-AG + MTL-ATG represented to the merged validation results of both models, and the MTL-AG-ATG for the integrated receptors with a merged of agonist and antagonist model using agonist datasets or antagonist datasets validation. The y-axis is the validation performance of the mean MSE for each GPCR