Skip to main content
Fig. 10 | Journal of Cheminformatics

Fig. 10

From: Graph neural processes for molecules: an evaluation on docking scores and strategies to improve generalization

Fig. 10

Bayesian optimization of druglike F2 (left) and selective JAK2 (right). MG-CNPs always reached either the best or a near-best molecule, outperforming ADKF-IFT and GPs with a Tanimoto kernel on binary FPs. Trajectory lines indicate the mean of 10 trajectories with different random initializations, and shaded area indicates the standard error. Horizontal dotted lines indicate the best and second-best molecules in the library of 60k compounds

Back to article page

Journal of Cheminformatics

ISSN: 1758-2946

Contact us