Structure-based drug design with geometric deep learning

Isert, Clemens; Atz, Kenneth; Schneider, Gisbert

doi:10.1016/j.sbi.2023.102548

Cited by 70 publications

(61 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We also demonstrated the value of EGGNet in improving results from blind docking. (1) , E (2) , E int ]) 0.583 0.638 W(BatchN orm([E (1) , E (2) , E int ])) 0.442 0.677 W(concat([E (1) , E (2) , E int ])) 0.474 0.694 W(BatchN orm(concat([E (1) , E (2) , E int ]))) 0.378 0.670…”

Section: Discussionmentioning

confidence: 99%

“…However, our results on the PDBbind/CASF-2016 experiments do not support this hypothesis (Table A1). This is probably due to the inaccurate estimation of the protein's energy E (1) because we had to use the binding pocket rather than the entire protein structure to construct the GoG to make the training memory efficient.…”

Section: Protein Small Molecule Binding Affinity Regressionmentioning

confidence: 99%

“…The ability to modulate PPIs with small molecules or synthetic peptides is a component of therapeutics development. There exist four classes of problems in structure-based drug design for which molecular interactions can be tackled by machine learning (ML) [1], including (i) protein complex property prediction, (ii) binding site/interface identification, (iii) docking (binding pose generation), and (iv) de novo design. Property prediction for protein complexes is a task that can also serve as an integral component for other tasks, such as evaluating binding poses generated by docking algorithms or predicting the affinity for computationally designed novel drug candidates.…”

Section: Introductionmentioning

confidence: 99%

“…We also desire the representation to be capable of handling synthetic peptides composed of both natural and non-canonical amino acids. 3D graphs are commonly used to represent the 3D structures of individual proteins and protein-small molecule complexes [1]. To increase its generalizbility and scalability, we extend the 3D graph to 3D graph of graphs (GoG) representation.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

EGGNet, a generalizable geometric deep learning framework for protein complex pose scoring

Wang

Brand

Adolf-Bryfogle

et al. 2023

Preprint

View full text Add to dashboard Cite

Computational prediction of molecule-protein interactions has been key for developing new molecules to interact with a target protein for therapeutics development. Literature includes two independent streams of approaches: (1) predicting protein-protein interactions between naturally occurring proteins and (2) predicting the binding affinities between proteins and small molecule ligands (aka drug target interaction, or DTI). Studying the two problems in isolation has limited computational models' ability to generalize across tasks, both of which ultimately involve non- covalent interactions with a protein target. In this work, we developed Equivariant Graph of Graphs neural Network (EGGNet), a geometric deep learning framework for molecule-protein binding predictions that can handle three types of molecules for interacting with a target protein: (1) small molecules, (2) synthetic peptides and (3) natural proteins. EGGNet leverages a graph of graphs (GoGs) representation constructed from the molecule structures at atomic-resolution and utilizes a multi-resolution equivariant graph neural network (GNN) to learn from such representations. In addition, EGGNet gets inspired by biophysics and makes use of both atom- and residue-level interactions, which greatly improve EGGNet's ability to rank candidate poses from blind docking. EGGNet achieves competitive performance on both a public protein-small molecule binding affinity prediction task (80.2% top-1 success rate on CASF-2016) and an synthetic protein interface prediction task (88.4% AUPR). We envision that the proposed geometric deep learning framework can generalize to many other protein interaction prediction problems, such as binding site prediction and molecular docking, helping to accelerate protein engineering and structure-based drug development.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Protein Small Molecule Binding Affinity Regressionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

EGGNet, a generalizable geometric deep learning framework for protein complex pose scoring

Wang

Brand

Adolf-Bryfogle

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Non-Euclidean convolutional neural networks (Monti et al, 2017) and point cloud-based learning models (Sverrisson et al, 2022) have been applied to encode the molecular surface for downstream applications, e.g., protein binding site prediction (Mylonas et al, 2021). However, existing methods apply filters with fixed sizes and are highly dependent on the surface mesh quality, which limit the expressive power for molecular shape representation across different spatial scales (Somnath et al, 2021;Isert et al, 2022).…”

Section: Related Workmentioning

confidence: 99%

Learning Harmonic Molecular Representations on Riemannian Manifold

Wang¹,

Shen²,

Chen³

et al. 2023

Preprint

View full text Add to dashboard Cite

Molecular representation learning plays a crucial role in AI-assisted drug discovery research. Encoding 3D molecular structures through Euclidean neural networks has become the prevailing method in the geometric deep learning community. However, the equivariance constraints and message passing in Euclidean space may limit the network expressive power. In this work, we propose a Harmonic Molecular Representation learning (HMR) framework, which represents a molecule using the Laplace-Beltrami eigenfunctions of its molecular surface. HMR offers a multi-resolution representation of molecular geometric and chemical features on 2D Riemannian manifold. We also introduce a harmonic message passing method to realize efficient spectral message passing over the surface manifold for better molecular encoding. Our proposed method shows comparable predictive power to current models in small molecule property prediction, and outperforms the state-of-the-art deep learning models for ligand-binding protein pocket classification and the rigid protein docking challenge, demonstrating its versatility in molecular representation learning.

show abstract

Template‐guided method for protein–ligand complex structure prediction: Application to CASP15 protein–ligand studies

Duan

Zou

2023

Proteins

View full text Add to dashboard Cite

Critical Assessment of Structure Prediction 15 (CASP15) added a new category of ligand prediction to promote the development of protein/RNA‐ligand modeling methods, which have become important tools in modern drug discovery. A total of 22 targets were released, including 18 protein–ligand targets and 4 RNA‐ligand targets. We applied our recently developed template‐guided method to the protein–ligand complex structure predictions. The method combined a physicochemical, molecular docking method, and a bioinformatics‐based ligand similarity method. The Protein Data Bank was scanned for template structures containing the target protein, homologous proteins, or proteins sharing a similar fold with the target protein. The binding modes of the co‐bound ligands in the template structures were used to guide the complex structure prediction for the target. The CASP assessment results show that the overall performance of our method was ranked second when the top predicted model was considered for each target. Here, we analyzed our predictions in detail, and discussed the challenges including protein conformational changes, large and flexible ligands, and multiple diverse ligands in a binding pocket.

show abstract

Structure-based drug design with geometric deep learning

Cited by 70 publications

References 53 publications

EGGNet, a generalizable geometric deep learning framework for protein complex pose scoring

EGGNet, a generalizable geometric deep learning framework for protein complex pose scoring

Learning Harmonic Molecular Representations on Riemannian Manifold

Template‐guided method for protein–ligand complex structure prediction: Application to CASP15 protein–ligand studies

Contact Info

Product

Resources

About