2019
DOI: 10.1021/acs.jcim.9b00387
|View full text |Cite
|
Sign up to set email alerts
|

Predicting Drug–Target Interaction Using a Novel Graph Neural Network with 3D Structure-Embedded Graph Representation

Abstract: Accurate prediction of drug-target interaction (DTI) is essential for in silico drug design. For the purpose, we propose a novel approach for predicting DTI using a GNN that directly incorporates the 3D structure of a protein-ligand complex. We also apply a distance-aware graph attention algorithm with gate augmentation to increase the performance of our model. As a result, our model shows better performance than docking and other deep learning methods for both virtual screening and pose prediction. In additio… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

6
342
1

Year Published

2019
2019
2024
2024

Publication Types

Select...
4
3
3

Relationship

0
10

Authors

Journals

citations
Cited by 332 publications
(349 citation statements)
references
References 56 publications
6
342
1
Order By: Relevance
“…We see similar magnitudes of improvement over our baseline docking results (reported in Table 2 in the main text) when compared with Ref. 18. In Table S6, we show that similar performance is also achieved after using the trained models reported in the main text (that is, models trained on all available poses) but removing poses between 2 Å and 4 Å RMSD from the test set.…”
Section: Binding Mode Prediction Results On "Filtered" Pdbind Datasetsupporting
confidence: 74%
“…We see similar magnitudes of improvement over our baseline docking results (reported in Table 2 in the main text) when compared with Ref. 18. In Table S6, we show that similar performance is also achieved after using the trained models reported in the main text (that is, models trained on all available poses) but removing poses between 2 Å and 4 Å RMSD from the test set.…”
Section: Binding Mode Prediction Results On "Filtered" Pdbind Datasetsupporting
confidence: 74%
“…Thus, for pose selection, precise comparisons between methods can only be made if the identical test set is used. As an example of the difficulty in comparing pose selection performance, consider the PDBbind-based evaluation of the graph-based model of Lim et al 26 . This model achieves an AUC of 0.968, which is higher than any of our PDBbind evaluations, but also exhibits a Top1 of less than 50%, which is substantial lower than our worse Top1 PDBbind statistic (77%, Figure 5).…”
Section: Discussionmentioning
confidence: 99%
“…Because ComBind can use any per-ligand docking method for pose generation and scoring of individual ligands, it will be able to take advantage of improvements to these methods. For example, several recent methods use machine learning to fit scoring functions (38)(39)(40), and others allow for binding pocket flexibility when generating candidate poses (8,36).…”
Section: Extensibility and Future Workmentioning
confidence: 99%