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

Dmitry S. Karlov, Sergey Sosnin, Maxim V. Fedorov^*, Petr Popov

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

82 Citations (Scopus)

52 Downloads (Pure)

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. ©

Original language	English
Pages (from-to)	5150-5159
Number of pages	10
Journal	ACS Omega
Volume	5
Issue number	10
Early online date	9 Mar 2020
DOIs	https://doi.org/10.1021/acsomega.9b04162
Publication status	Published - 17 Mar 2020

Funding

P.P. acknowledges Russian Science Foundation (RSF) research grant 18-74-00117. The results of the work were obtained using computational resources of MCC NRC ”Kurchatov Institute, http://computing.nrcki.ru/ .

Keywords

protein ligand scoring
protein molecules
protein data bank
neural networks

Access to Document

10.1021/acsomega.9b04162Licence: CC BY-NC-ND 4.0

Karlov-etal-ACSO-2020-graphDelta-MPNN-scoring-function-for-the-affinity-predictionFinal published version, 2.33 MBLicence: CC BY-NC-ND 4.0

Cite this

@article{884e2db271db44a191835e37fdebb9a7,

title = "GraphDelta: MPNN scoring function for the affinity prediction of protein-ligand complexes",

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 (Kd), inhibition constant (Ki), 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 Kdeep. {\textcopyright}",

keywords = "protein ligand scoring, protein molecules, protein data bank, neural networks",

author = "Karlov, \{Dmitry S.\} and Sergey Sosnin and Fedorov, \{Maxim V.\} and Petr Popov",

year = "2020",

month = mar,

day = "17",

doi = "10.1021/acsomega.9b04162",

language = "English",

volume = "5",

pages = "5150--5159",

journal = "ACS Omega",

issn = "2470-1343",

number = "10",

}

TY - JOUR

T1 - GraphDelta

T2 - MPNN scoring function for the affinity prediction of protein-ligand complexes

AU - Karlov, Dmitry S.

AU - Sosnin, Sergey

AU - Fedorov, Maxim V.

AU - Popov, Petr

PY - 2020/3/17

Y1 - 2020/3/17

N2 - 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 (Kd), inhibition constant (Ki), 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 Kdeep. ©

AB - 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 (Kd), inhibition constant (Ki), 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 Kdeep. ©

KW - protein ligand scoring

KW - protein molecules

KW - protein data bank

KW - neural networks

UR - https://www.scopus.com/pages/publications/85081686625

U2 - 10.1021/acsomega.9b04162

DO - 10.1021/acsomega.9b04162

M3 - Article

AN - SCOPUS:85081686625

SN - 2470-1343

VL - 5

SP - 5150

EP - 5159

JO - ACS Omega

JF - ACS Omega

IS - 10

ER -

GraphDelta: MPNN scoring function for the affinity prediction of protein-ligand complexes

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this