Improving the Accuracy of Protein-Ligand Binding Mode Prediction Using a Molecular Dynamics-Based Pocket Generation Approach

AAp-MSMD: Amino Acid Preference Mapping on Protein-Protein Interaction Surfaces Using Mixed-Solvent Molecular Dynamics

Peptides have attracted much attention recently owing to their well-balanced properties as drugs against protein-protein interaction (PPI) surfaces. Molecular simulation-based predictions of binding sites and amino acid residues with high affinity to PPI surfaces are expected to accelerate the design of peptide drugs. Mixed-solvent molecular dynamics (MSMD), which adds probe molecules or fragments of functional groups as solutes to the hydration model, detects the binding hotspots and cryptic sites induced by small molecules. The detection results vary depending on the type of probe molecule; thus, they provide important information for drug design. For rational peptide drug design using MSMD, we proposed MSMD with amino acid residue probes, named amino acid probe-based MSMD (AAp-MSMD), to detect hotspots and identify favorable amino acid types on protein surfaces to which peptide drugs bind. We assessed our method in terms of hotspot detection at the amino acid probe level and binding free energy prediction with amino acid probes at the PPI site for the complex structure that formed the PPI. In hotspot detection, the max-spatial probability distribution map (max-PMAP) obtained from AAp-MSMD detected the PPI site, to which each type of amino acid can bind favorably. In the binding free energy prediction using amino acid probes, ΔGFE obtained from AAp-MSMD roughly estimated the experimental binding affinities from the structure-activity relationship. AAp-MSMD, with amino acid probes, provides estimated binding sites and favorable amino acid types at the PPI site of a target protein.

Are Deep Learning Structural Models Sufficiently Accurate for Virtual Screening? Application of Docking Algorithms to AlphaFold2 Predicted Structures

Machine learning-based protein structure prediction algorithms, such as RosettaFold and AlphaFold2, have greatly impacted the structural biology field, arousing a fair amount of discussion around their potential role in drug discovery. While there are few preliminary studies addressing the usage of these models in virtual screening, none of them focus on the prospect of hit-finding in a real-world virtual screen with a model based on low prior structural information. In order to address this, we have developed an AlphaFold2 version where we exclude all structural templates with more than 30% sequence identity from the model-building process. In a previous study, we used those models in conjunction with state-of-the-art free energy perturbation methods and demonstrated that it is possible to obtain quantitatively accurate results. In this work, we focus on using these structures in rigid receptor-ligand docking studies. Our results indicate that using out-of-the-box Alphafold2 models is not an ideal scenario for virtual screening campaigns; in fact, we strongly recommend to include some post-processing modeling to drive the binding site into a more realistic holo model.

GraphSite: Ligand Binding Site Classification with Deep Graph Learning

The binding of small organic molecules to protein targets is fundamental to a wide array of cellular functions. It is also routinely exploited to develop new therapeutic strategies against a variety of diseases. On that account, the ability to effectively detect and classify ligand binding sites in proteins is of paramount importance to modern structure-based drug discovery. These complex and non-trivial tasks require sophisticated algorithms from the field of artificial intelligence to achieve a high prediction accuracy. In this communication, we describe GraphSite, a deep learning-based method utilizing a graph representation of local protein structures and a state-of-the-art graph neural network to classify ligand binding sites. Using neural weighted message passing layers to effectively capture the structural, physicochemical, and evolutionary characteristics of binding pockets mitigates model overfitting and improves the classification accuracy. Indeed, comprehensive cross-validation benchmarks against a large dataset of binding pockets belonging to 14 diverse functional classes demonstrate that GraphSite yields the class-weighted F1-score of 81.7%, outperforming other approaches such as molecular docking and binding site matching. Further, it also generalizes well to unseen data with the F1-score of 70.7%, which is the expected performance in real-world applications. We also discuss new directions to improve and extend GraphSite in the future.

On the vibrational free energy of hydrated proteins

When the hydration shell of a protein is filled with at least 0.6 gram of water per gram of protein, a significant anti-correlation between the vibrational free energy and the potential energy of energy-minimized conformers is observed. This means that low potential energy, well-hydrated, protein conformers tend to be more rigid than high-energy ones. On the other hand, in the case of CASP target 624, when its hydration shell is filled, a significant energy gap is observed between the crystal structure and the best conformers proposed during the prediction experiment, strongly suggesting that including explicit water molecules may help identifying unlikely conformers among good-looking ones.

Higher Accuracy Achieved for Protein-Ligand Binding Pose Prediction by Elastic Network Model-Based Ensemble Docking

Molecular docking plays an indispensable role in predicting the receptor-ligand interactions in which the protein receptor is usually kept rigid, whereas the ligand is treated as being flexible. Because of the inherent flexibility of proteins, the binding pocket of apo receptors might undergo significant conformational rearrangement upon ligand binding, which limits the prediction accuracy of docking. Here, we present an iterative anisotropic network model (iterANM)-based ensemble docking approach, which generates multiple holo-like receptor structures starting from the apo receptor and incorporates protein flexibility into docking. In a validation data set consisting of 233 chemically diverse cyclin-dependent kinase 2 (CDK2) inhibitors, the iterANM-based ensemble docking achieves higher capacity to reproduce native-like binding poses compared with those using single apo receptor conformation or conformational ensemble from molecular dynamics simulations. The prediction success rate within the top5-ranked binding poses produced by the iterANM can further be improved through reranking with the molecular mechanics-Poisson-Boltzmann surface area method. In a smaller data set with 58 CDK2 inhibitors, the iterANM-based ensemble shows a higher success rate compared with the flexible receptor-based docking procedure AutoDockFR and other receptor conformation generation approaches. Further, an additional docking test consisting of 10 diverse receptor-ligand combinations shows that the iterANM is robustly applicable for different receptor structures. These results suggest the iterANM-based ensemble docking as an accurate, efficient, and practical framework to predict the binding mode of a ligand for receptors with flexibility.

BionoiNet: ligand-binding site classification with off-the-shelf deep neural network

MOTIVATION

Fast and accurate classification of ligand-binding sites in proteins with respect to the class of binding molecules is invaluable not only to the automatic functional annotation of large datasets of protein structures but also to projects in protein evolution, protein engineering and drug development. Deep learning techniques, which have already been successfully applied to address challenging problems across various fields, are inherently suitable to classify ligand-binding pockets. Our goal is to demonstrate that off-the-shelf deep learning models can be employed with minimum development effort to recognize nucleotide- and heme-binding sites with a comparable accuracy to highly specialized, voxel-based methods.

RESULTS

We developed BionoiNet, a new deep learning-based framework implementing a popular ResNet model for image classification. BionoiNet first transforms the molecular structures of ligand-binding sites to 2D Voronoi diagrams, which are then used as the input to a pretrained convolutional neural network classifier. The ResNet model generalizes well to unseen data achieving the accuracy of 85.6% for nucleotide- and 91.3% for heme-binding pockets. BionoiNet also computes significance scores of pocket atoms, called BionoiScores, to provide meaningful insights into their interactions with ligand molecules. BionoiNet is a lightweight alternative to computationally expensive 3D architectures.

AVAILABILITY AND IMPLEMENTATION

BionoiNet is implemented in Python with the source code freely available at: https://github.com/CSBG-LSU/BionoiNet.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.