graphDelta: MPNN Scoring Function for the Affinity Prediction of Protein-Ligand Complexes

Dmitry S Karlov; Sergey Sosnin; Maxim V Fedorov; Petr Popov

doi:10.1021/acsomega.9b04162

graphDelta: MPNN Scoring Function for the Affinity Prediction of Protein-Ligand Complexes

ACS Omega. 2020 Mar 9;5(10):5150-5159. doi: 10.1021/acsomega.9b04162. eCollection 2020 Mar 17.

Authors

Dmitry S Karlov¹, Sergey Sosnin^{1

2}, Maxim V Fedorov^{1

2

3}, Petr Popov^{1

4}

Affiliations

¹ Skolkovo Institute of Science and Technology, Moscow 143026, Russia.
² Skolkovo Innovation Center,Syntelly LLC, 42 Bolshoy Boulevard, Moscow 143026, Russia.
³ University of Strathclyde, Physics John Anderson Building, 107 Rottenrow East, Glasgow UK G4 0NG, U.K.
⁴ Moscow Institute of Physics and Technology, Dolgoprudny 141701, Russia.

Abstract

In this work, we present graph-convolutional neural networks for the prediction of binding constants of protein-ligand complexes. We derived the model using multi task learning, where the target variables are the dissociation constant (K _d), inhibition constant (K _i), and half maximal inhibitory concentration (IC₅₀). Being rigorously trained on the PDBbind dataset, the model achieves the Pearson correlation coefficient of 0.87 and the RMSE value of 1.05 in pK units, outperforming recently developed 3D convolutional neural network model K _deep.