From Static to Dynamic Structures: Improving Binding Affinity Prediction with Graph-Based Deep Learning

Yaosen Min,Ye Wei,Peizhuo Wang,Xiaoting Wang,Han Li,Nian Wu,Stefan Bauer,Shuxin Zheng,Yu Shi,Yingheng Wang,Ji Wu,Dan Zhao,Jianyang Zeng

DOI: https://doi.org/10.1002/advs.202405404

2024-09-02

Abstract:Accurate prediction of protein-ligand binding affinities is an essential challenge in structure-based drug design. Despite recent advances in data-driven methods for affinity prediction, their accuracy is still limited, partially because they only take advantage of static crystal structures while the actual binding affinities are generally determined by the thermodynamic ensembles between proteins and ligands. One effective way to approximate such a thermodynamic ensemble is to use molecular dynamics (MD) simulation. Here, an MD dataset containing 3,218 different protein-ligand complexes is curated, and Dynaformer, a graph-based deep learning model is further developed to predict the binding affinities by learning the geometric characteristics of the protein-ligand interactions from the MD trajectories. In silico experiments demonstrated that the model exhibits state-of-the-art scoring and ranking power on the CASF-2016 benchmark dataset, outperforming the methods hitherto reported. Moreover, in a virtual screening on heat shock protein 90 (HSP90) using Dynaformer, 20 candidates are identified and their binding affinities are further experimentally validated. Dynaformer displayed promising results in virtual drug screening, revealing 12 hit compounds (two are in the submicromolar range), including several novel scaffolds. Overall, these results demonstrated that the approach offer a promising avenue for accelerating the early drug discovery process.

Biomolecules,Machine Learning,Chemical Physics,Quantitative Methods

What problem does this paper attempt to address?

The paper aims to address the problem of protein-ligand binding affinity prediction, which is a key challenge in drug design. Despite recent advances in data-driven approaches, the accuracy of these methods remains limited, partly because they only utilize static crystal structures, whereas actual binding affinity is often determined by the thermodynamic ensemble of protein-ligand interactions. To tackle this issue, the research team developed a large-scale dataset based on molecular dynamics (MD) simulations and proposed a graph neural network model named Dynaformer. Dynaformer predicts binding affinity by learning the geometric features of protein-ligand interactions from MD trajectories. Experimental results show that the model achieved state-of-the-art performance in scoring and ranking on the CASF-2016 benchmark dataset. Additionally, in a virtual screening against Heat Shock Protein 90 (HSP90), Dynaformer successfully identified 20 candidate compounds, and the binding affinities of these compounds were experimentally validated. These results demonstrate the potential of Dynaformer in the early drug discovery process, particularly in accelerating the design of novel drugs.

From Static to Dynamic Structures: Improving Binding Affinity Prediction with Graph-Based Deep Learning

From Static to Dynamic Structures: Improving Binding Affinity Prediction with a Graph-Based Deep Learning Model

From Static to Dynamic Structures: Improving Binding Affinity Prediction with Graph‐Based Deep Learning

Predicting the Protein-Ligand Affinity from Molecular Dynamics Trajectories

DynamicBind: predicting ligand-specific protein-ligand complex structure with a deep equivariant generative model

Can molecular dynamics simulations improve predictions of protein-ligand binding affinity with machine learning?

An Adaptive Graph Learning Method for Automated Molecular Interactions and Properties Predictions

Improved Protein–Ligand Binding Affinity Prediction with Structure-Based Deep Fusion Inference

Development and evaluation of a deep learning model for protein-ligand binding affinity prediction

Deep Scoring Neural Network Replacing the Scoring Function Components to Improve the Performance of Structure-Based Molecular Docking.

DLSSAffinity: Protein-Ligand Binding Affinity Prediction Via a Deep Learning Model

DEAttentionDTA: protein–ligand binding affinity prediction based on dynamic embedding and self-attention

Enhancing Drug-Target Binding Affinity Prediction through Deep Learning and Protein Secondary Structure Integration

Binding Affinity Prediction with 3D Machine Learning: Training Data and Challenging External Testing

Structure-based, deep-learning models for protein-ligand binding affinity prediction

Assessment of molecular dynamics time series descriptors in protein-ligand affinity prediction.

Exploring protein–ligand binding affinity prediction with electron density-based geometric deep learning

Accurate prediction of protein–ligand interactions by combining physical energy functions and graph-neural networks

Leveraging nonstructural data to predict structures and affinities of protein-ligand complexes

Fusion-Based Deep Learning Architecture for Detecting Drug-Target Binding Affinity Using Target and Drug Sequence and Structure.

DGDTA: dynamic graph attention network for predicting drug-target binding affinity