Fig. 3

Incorporating information about binding assay conditions improves protein-protein affinity prediction. a We plot the Pearson correlation between predicted and experimentally-determined binding affinities (pKd) using models with (white) and without (gray) three features describing the conditions under which binding affinities were measured experimentally (temperature, pH and the assay method). b We plot the correlation between predicted and experimentally-determined binding affinities for the same models examined in (a), using unseen testing data generated by leave-one-out cross validation (see Methods). c We plot the RMSD between predicted and experimentally-determined binding affinities for the same models in (a and b), using leave-one-out cross-validation. In each panel, bars indicate standard errors