DPLA: prediction of protein-ligand binding affinity by integrating multi-level information

Protein-ligand interactions are the molecular basis of many important cellular activities, such as gene regulation, cell metabolism, and signal transduction. Protein-ligand binding affinity is a crucial metric of the strength of the interaction between the two, and accurate prediction of its binding affinity is essential for discovering drugs' new uses. So far, although many predictive models based on machine learning and deep learning have been reported, most of the models mainly focus on one-dimensional sequence and two-dimensional structural characteristics of proteins and ligands, but fail to deeply explore the detailed interaction information between proteins and ligand atoms in the binding pocket region of three-dimensional space. In this study, we introduced a novel 4D tensor feature to capture key interactions within the binding pocket and developed a three-dimensional convolutional neural network (CNN) model based on this feature. Using ten-fold cross-validation, we identified the optimal parameter combination and pocket size. Additionally, we employed feature engineering to extract features across multiple dimensions, including one-dimensional sequences, two-dimensional structures of the ligand and protein, and three-dimensional interaction features between them. We proposed an efficient protein-ligand binding affinity prediction model MCDTA (multi-dimensional convolutional drug-target affinity), built on a multi-dimensional convolutional neural network framework. Feature ablation experiments revealed that the 4D tensor feature had the most significant impact on model performance. MCDTA performed exceptionally well on the PDBbind v.2020 dataset, achieving an RMSE of 1.231 and a PCC of 0.823. In comparative experiments, it outperformed five other mainstream binding affinity prediction models, with an RMSE of 1.349 and a PCC of 0.795. Moreover, MCDTA demonstrated strong generalization ability and practical screening performance across multiple benchmark datasets, highlighting its reliability and accuracy in predicting protein-ligand binding affinity. The code for MCDTA is available at https://github.com/dfhuang-AI/MCDTA .

PLAGCA: Predicting protein-ligand binding affinity with the graph cross-attention mechanism.

Inductive bias in machine learning (ML) is the set of assumptions describing how a model makes predictions. Different ML-based methods for protein-ligand binding affinity (PLA) prediction have different inductive biases, leading to different levels of generalization capability and interpretability. Intuitively, the inductive bias of an ML-based model for PLA prediction should fit in with biological mechanisms relevant for binding to achieve good predictions with meaningful reasons. To this end, we propose an interaction-based inductive bias to restrict neural networks to functions relevant for binding with two assumptions: 1) A protein-ligand complex can be naturally expressed as a heterogeneous graph with covalent and non-covalent interactions; 2) The predicted PLA is the sum of pairwise atom-atom affinities determined by non-covalent interactions. The interaction-based inductive bias is embodied by an explainable heterogeneous interaction graph neural network (EHIGN) for explicitly modeling pairwise atom-atom interactions to predict PLA from 3D structures. Extensive experiments demonstrate that EHIGN achieves better generalization capability than other state-of-the-art ML-based baselines in PLA prediction and structure-based virtual screening. More importantly, comprehensive analyses of distance-affinity, pose-affinity, and substructure-affinity relations suggest that the interaction-based inductive bias can guide the model to learn atomic interactions that are consistent with physical reality. As a case study to demonstrate practical usefulness, our method is tested for predicting the efficacy of Nirmatrelvir against SARS-CoV-2 variants. EHIGN successfully recognizes the changes in the efficacy of Nirmatrelvir for different SARS-CoV-2 variants with meaningful reasons.

Accurate prediction of protein-ligand binding affinity is pivotal for drug design and discovery. Here, we proposed a novel deep fusion graph neural networks framework named FGNN to learn the protein-ligand interactions from the 3D structures of protein-ligand complexes. Unlike 1D sequences for proteins or 2D graphs for ligands, the 3D graph of protein-ligand complex enables the more accurate representations of the protein-ligand interactions. Benchmark studies have shown that our fusion models FGNN can achieve more accurate prediction of binding affinity than any individual algorithm. The advantages of fusion strategies have been demonstrated in terms of expressive power of data, learning efficiency and model interpretability. Our fusion models show satisfactory performances on diverse data sets, demonstrating their generalization ability. Given the good performances in both binding affinity prediction and virtual screening, our fusion models are expected to be practically applied for drug screening and design. Our work highlights the potential of the fusion graph neural network algorithm in solving complex prediction problems in computational biology and chemistry. The fusion graph neural networks (FGNN) model is freely available in https://github.com/LinaDongXMU/FGNN.

Pharmaceutical Applications of the Polarizable Amoeba Potential, Including Protein-Ligand Binding Affinity and Drug Solubility, using the Force Field X Software

Predicting protein-ligand binding affinity is a central issue in drug design. Various deep learning models have been published in recent years, where many of them rely on 3D protein-ligand complex structures as input and tend to focus on the single task of reproducing binding affinity. In this study, we have developed a graph neural network model called PLANET (Protein-Ligand Affinity prediction NETwork). This model takes the graph-represented 3D structure of the binding pocket on the target protein and the 2D chemical structure of the ligand molecule as input. It was trained through a multi-objective process with three related tasks, including deriving the protein-ligand binding affinity, protein-ligand contact map, and ligand distance matrix. Besides the protein-ligand complexes with known binding affinity data retrieved from the PDBbind database, a large number of non-binder decoys were also added to the training data for deriving the final model of PLANET. When tested on the CASF-2016 benchmark, PLANET exhibited a scoring power comparable to the best result yielded by other deep learning models as well as a reasonable ranking power and docking power. In virtual screening trials conducted on the DUD-E benchmark, PLANET's performance was notably better than several deep learning and machine learning models. As on the LIT-PCBA benchmark, PLANET achieved comparable accuracy as the conventional docking program Glide, but it only spent less than 1% of Glide's computation time to finish the same job because PLANET did not need exhaustive conformational sampling. Considering the decent accuracy and efficiency of PLANET in binding affinity prediction, it may become a useful tool for conducting large-scale virtual screening.

Accurately predicting protein-ligand binding affinity is an important step in the drug discovery process. Deep learning (DL) methods have improved binding affinity prediction by using diverse categories of molecular data. However, many models rely heavily on interaction and sequence data, which impedes proper learning and limits performance in de novo applications. To address these limitations, we developed a novel graph neural network model, called StructureNet (structure-based graph neural network), to predict protein-ligand binding affinity. StructureNet improves existing DL methods by focusing entirely on structural descriptors to mitigate data memorization issues introduced by sequence and interaction data. StructureNet represents the protein and ligand structures as graphs, which are processed using a GNN-based ensemble deep learning model. StructureNet achieved a PCC of 0.68 and an AUC of 0.75 on the PDBBind v.2020 Refined Set, outperforming similar structure-based models. External validation on the DUDE-Z dataset showed that StructureNet can effectively distinguish between active and decoy ligands. Further testing on a small subset of well-known drugs indicates that StructureNet has high potential for rapid virtual screening applications. We also hybridized StructureNet with interaction- and sequence-based models to investigate their impact on testing accuracy and found minimal difference (0.01 PCC) between merged models and StructureNet as a standalone model. An ablation study found that geometric descriptors were the key drivers of model performance, with their removal leading to a PCC decrease of over 15.7%. Lastly, we tested StructureNet on ensembles of binding complex conformers generated using molecular dynamics (MD) simulations and found that incorporating multiple conformations of the same complex often improves model accuracy by capturing binding site flexibility. Overall, the results show that structural data alone are sufficient for binding affinity predictions and can address pattern recognition challenges introduced by sequence and interaction features. Additionally, structural representations of protein-ligand complexes can be considerably improved using geometric and topological descriptors. We made StructureNet GUI interface freely available online.

Geometric graph learning with extended atom-types features for protein-ligand binding affinity prediction

Structure-guided drug design efforts stand to benefit greatly from insights provided by accurate predictive models of binding affinity. Here we use Delaunay tessellation to analyze atomic four-body interactions in a diverse set of high-resolution crystal structures for three hundred protein-ligand complexes, each with an experimentally determined dissociation constant. Two hundred of these complexes are randomly selected to derive an empirical function for predicting the standard Gibbs free energy of binding (ΔG). In each case, structural atomic coordinates are used to generate tessellations for the protein-ligand complex as well as for the isolated protein without the bound ligand. An all-atom four-body knowledge-based statistical potential is used to generate a topological score based on each tessellation, and binding affinity is calculated as an empirical linear function of the difference between the topological scores of the complex and the isolated protein. Experimental and calculated ΔG values are in good agreement, with a correlation coefficient (r) of 0.79 and a standard error of 1.98 kcal/mol. To evaluate the method, tessellation-based topological score differences are calculated for the remaining one hundred complexes, and the linear transformation derived from the training data is used to predict their binding affinities. Again, the correlation coefficient is 0.79 between experimental and calculated binding energies for the test set, with a standard error of 1.93 kcal/mol.

Accurate prediction of protein-ligand bound poses and their affinities is essential in structure-based drug discovery. Here, we present an integrated deep-learning framework that disentangles the two core tasks─affinity estimation and pose evaluation─within complementary graph neural network architectures. BA-Pred is for predicting binding affinity and RMSD-Pred is for binding pose assessment, predicting the root-mean-squared deviation of ligand poses from crystal structures. Both models employ a Gated Graph Convolutional Network with Learnable Structural Positional Encoding (GatedGCN-LSPE) architecture to capture spatial and chemical dependencies across protein-ligand graphs. BA-Pred achieved state-of-the-art scoring power on the CASF-2016 benchmark with a root-mean-squared error of 1.10 pKd, while RMSD-Pred exhibited strong docking power with a top-1 success rate of 96%, comparable to the best reported deep-learning scoring functions. The robust generalization capability of RMSD-Pred was further validated on the external Astex diverse set and PoseBusters benchmarks, where it significantly improved the pose selection success rates of AutoDock-GPU by up to 33.1%. The accuracy of our methodology was demonstrated on pharmaceutical targets in the 16th Critical Assessment of Structure Prediction, where our approach ranked second in the ligand binding affinity prediction category. By using our models, an integrated pipeline was developed for virtual screening, where pose selection was performed with RMSD-Pred and binding affinities were predicted with BA-Pred. This combined approach demonstrated robust screening performance, achieving an enrichment factor (EF) 1% of 21.1 on the CASF-2016 benchmark. Furthermore, on the LIT-PCBA benchmark, rescoring poses docked by AutoDock-GPU with our pipeline significantly improved the EF 1% from 2.18 to 3.19. These various benchmark results demonstrate that our graph-neural network models show good and balanced performance in diverse protein-ligand interaction prediction tasks. Thus, we expect that our models will serve as a promising framework to accelerate the drug discovery process.

The recently released Boltz-2 cofolding model is generating high expectations by enabling both protein-ligand structure and binding affinity predictions. When applied to a recently described and challenging data set of ultralarge-virtual-screening hits, Boltz-2 excels at discriminating true from false positives, overcoming by a large margin all scoring functions tested so far on raw docking poses. Strikingly, affinity predictions seem to be relatively independent of pose quality but are not biased by obvious chemical similarity to known compounds sharing comparable binding potencies. To ascertain that Boltz-2 truly relies on the physics of intermolecular interactions, we challenged affinity predictions with biologically meaningful challenges (target mutation and target shuffling). Binary classification of active vs inactive compounds remains insensitive to key binding site mutations and even in some cases to target exchange, raising concerns on the hidden features governing Boltz-2 affinity predictions.

Accurate and rapid prediction of protein-ligand binding affinities is critical for drug discovery, particularly when evaluating large chemical libraries or new drug molecules from high-throughput generative models. We present UCBbind, a hybrid framework that combines a similarity-based transfer module with a deep learning-based prediction module to efficiently estimate binding affinities of small molecules to target proteins. For each query protein-ligand pair, UCBbind transfers experimental data from highly similar reference pairs when available, and applies the prediction module when no sufficiently similar reference exists. We benchmarked UCBbind on multiple datasets, including the PDBbind v2020 core set, the HiQBind dataset post 2020, and the COVID Moonshot database. Our results show that UCBbind achieves state-of-the-art predictive performance, particularly for test entries with high similarity to well-characterized reference proteins and ligands, and can support downstream tasks such as binding site prediction and binder/non-binder classification.

Predicting protein-ligand binding affinity is a crucial and challenging task in structure-based drug discovery. With the accumulation of complex structures and binding affinity data, various machine-learning scoring functions, particularly those based on deep learning, have been developed for this task, exhibiting superiority over their traditional counterparts. A fusion model sequentially connecting a graph neural network (GNN) and a convolutional neural network (CNN) to predict protein-ligand binding affinity is proposed in this work. In this model, the intermediate outputs of the GNN layers, as supplementary descriptors of atomic chemical environments at different levels, are concatenated with the input features of CNN. The model demonstrates a noticeable improvement in performance on CASF-2016 benchmark compared to its constituent CNN models. The generalization ability of the model is evaluated by setting a series of thresholds for ligand extended-connectivity fingerprint similarity or protein sequence similarity between the training and test sets. Masking experiment reveals that model can capture key interaction regions. Furthermore, the fusion model is applied to a virtual screening task for a novel target, PI5P4Kα. The fusion strategy significantly improves the ability of the constituent CNN model to identify active compounds. This work offers a novel approach to enhancing the accuracy of deep learning models in predicting binding affinity through fusion strategies.

Predicting accurate protein-ligand binding affinities is an important task in drug discovery but remains a challenge even with computationally expensive biophysics-based energy scoring methods and state-of-the-art deep learning approaches. Despite the recent advances in the application of deep convolutional and graph neural network-based approaches, it remains unclear what the relative advantages of each approach are and how they compare with physics-based methodologies that have found more mainstream success in virtual screening pipelines. We present fusion models that combine features and inference from complementary representations to improve binding affinity prediction. This, to our knowledge, is the first comprehensive study that uses a common series of evaluations to directly compare the performance of three-dimensional (3D)-convolutional neural networks (3D-CNNs), spatial graph neural networks (SG-CNNs), and their fusion. We use temporal and structure-based splits to assess performance on novel protein targets. To test the practical applicability of our models, we examine their performance in cases that assume that the crystal structure is not available. In these cases, binding free energies are predicted using docking pose coordinates as the inputs to each model. In addition, we compare these deep learning approaches to predictions based on docking scores and molecular mechanic/generalized Born surface area (MM/GBSA) calculations. Our results show that the fusion models make more accurate predictions than their constituent neural network models as well as docking scoring and MM/GBSA rescoring, with the benefit of greater computational efficiency than the MM/GBSA method. Finally, we provide the code to reproduce our results and the parameter files of the trained models used in this work. The software is available as open source at https://github.com/llnl/fast. Model parameter files are available at ftp://gdo-bioinformatics.ucllnl.org/fast/pdbbind2016_model_checkpoints/.

The ability to accurately predict the interaction of a ligand with its receptor is a key limitation in computer-aided drug design approaches such as virtual screening and de novo design. In this article, we examine current strategies for a physics-based approach to scoring of protein-ligand affinity, as well as outlining recent developments in force fields and quantum chemical techniques. We also consider advances in the development and application of simulation-based free energy methods to study protein-ligand interactions. Fuelled by recent advances in computational algorithms and hardware, there is the opportunity for increased integration of physics-based scoring approaches at earlier stages in computationally guided drug discovery. Specifically, we envisage increased use of implicit solvent models and simulation-based scoring methods as tools for computing the affinities of large virtual ligand libraries. Approaches based on end point simulations and reference potentials allow the application of more advanced potential energy functions to prediction of protein-ligand binding affinities. Comprehensive evaluation of polarizable force fields and quantum mechanical (QM)/molecular mechanical and QM methods in scoring of protein-ligand interactions is required, particularly in their ability to address challenging targets such as metalloproteins and other proteins that make highly polar interactions. Finally, we anticipate increasingly quantitative free energy perturbation and thermodynamic integration methods that are practical for optimization of hits obtained from screened ligand libraries.