Representability of algebraic topology for biomolecules in machine learning based scoring and virtual screening

Multiphysical graph neural network (MP-GNN) for COVID-19 drug design

Graph neural networks (GNNs) are the most promising deep learning models that can revolutionize non-Euclidean data analysis. However, their full potential is severely curtailed by poorly represented molecular graphs and features. Here, we propose a multiphysical graph neural network (MP-GNN) model based on the developed multiphysical molecular graph representation and featurization. All kinds of molecular interactions, between different atom types and at different scales, are systematically represented by a series of scale-specific and element-specific graphs with distance-related node features. From these graphs, graph convolution network (GCN) models are constructed with specially designed weight-sharing architectures. Base learners are constructed from GCN models from different elements at different scales, and further consolidated together using both one-scale and multi-scale ensemble learning schemes. Our MP-GNN has two distinct properties. First, our MP-GNN incorporates multiscale interactions using more than one molecular graph. Atomic interactions from various different scales are not modeled by one specific graph (as in traditional GNNs), instead they are represented by a series of graphs at different scales. Second, it is free from the complicated feature generation process as in conventional GNN methods. In our MP-GNN, various atom interactions are embedded into element-specific graph representations with only distance-related node features. A unique GNN architecture is designed to incorporate all the information into a consolidated model. Our MP-GNN has been extensively validated on the widely used benchmark test datasets from PDBbind, including PDBbind-v2007, PDBbind-v2013 and PDBbind-v2016. Our model can outperform all existing models as far as we know. Further, our MP-GNN is used in coronavirus disease 2019 drug design. Based on a dataset with 185 complexes of inhibitors for severe acute respiratory syndrome coronavirus (SARS-CoV/SARS-CoV-2), we evaluate their binding affinities using our MP-GNN. It has been found that our MP-GNN is of high accuracy. This demonstrates the great potential of our MP-GNN for the screening of potential drugs for SARS-CoV-2. Availability: The Multiphysical graph neural network (MP-GNN) model can be found in https://github.com/Alibaba-DAMO-DrugAI/MGNN. Additional data or code will be available upon reasonable request.

Hodge theory-based biomolecular data analysis

Hodge theory reveals the deep intrinsic relations of differential forms and provides a bridge between differential geometry, algebraic topology, and functional analysis. Here we use Hodge Laplacian and Hodge decomposition models to analyze biomolecular structures. Different from traditional graph-based methods, biomolecular structures are represented as simplicial complexes, which can be viewed as a generalization of graph models to their higher-dimensional counterparts. Hodge Laplacian matrices at different dimensions can be generated from the simplicial complex. The spectral information of these matrices can be used to study intrinsic topological information of biomolecular structures. Essentially, the number (or multiplicity) of k-th dimensional zero eigenvalues is equivalent to the k-th Betti number, i.e., the number of k-th dimensional homology groups. The associated eigenvectors indicate the homological generators, i.e., circles or holes within the molecular-based simplicial complex. Furthermore, Hodge decomposition-based HodgeRank model is used to characterize the folding or compactness of the molecular structures, in particular, the topological associated domain (TAD) in high-throughput chromosome conformation capture (Hi-C) data. Mathematically, molecular structures are represented in simplicial complexes with certain edge flows. The HodgeRank-based average/total inconsistency (AI/TI) is used for the quantitative measurements of the folding or compactness of TADs. This is the first quantitative measurement for TAD regions, as far as we know.

Machine-Learning- and Knowledge-Based Scoring Functions Incorporating Ligand and Protein Fingerprints

We propose a novel machine-learning-based scoring function for drug discovery that incorporates ligand and protein structural information into a knowledge-based PMF score. Molecular docking, a simulation method for structure-based drug design (SBDD), is expected to reduce the enormous costs associated with conventional experimental methods in terms of rational drug discovery. Molecular docking has two main purposes: to predict ligand-binding structures for target proteins and to predict protein-ligand binding affinity. Currently available programs of molecular docking offer an accurate prediction of ligand binding structures for many systems. However, the accurate prediction of binding affinity remains challenging. In this study, we developed a new scoring function that incorporates fingerprints representing ligand and protein structures as descriptors in the PMF score. Here, regression analysis of the scoring function was performed using the following machine learning techniques: least absolute shrinkage and selection operator (LASSO) and light gradient boosting machine (LightGBM). The results on a test data set showed that the binding affinity delivered by the newly developed scoring function has a Pearson correlation coefficient of 0.79 with the experimental value, which surpasses that of the conventional scoring functions. Further analysis provided a chemical understanding of the descriptors that contributed significantly to the improvement in prediction accuracy. Our approach and findings are useful for rational drug discovery.

ASPECTS OF TOPOLOGICAL APPROACHES FOR DATA SCIENCE

We establish a new theory which unifies various aspects of topological approaches for data science, by being applicable both to point cloud data and to graph data, including networks beyond pairwise interactions. We generalize simplicial complexes and hypergraphs to super-hypergraphs and establish super-hypergraph homology as an extension of simplicial homology. Driven by applications, we also introduce super-persistent homology.

Differences in ligand-induced protein dynamics extracted from an unsupervised deep learning approach correlate with protein-ligand binding affinities

Prediction of protein–ligand binding affinity is a major goal in drug discovery. Generally, free energy gap is calculated between two states (e.g., ligand binding and unbinding). The energy gap implicitly includes the effects of changes in protein dynamics induced by ligand binding. However, the relationship between protein dynamics and binding affinity remains unclear. Here, we propose a method that represents ligand-binding-induced protein behavioral change with a simple feature that can be used to predict protein–ligand affinity. From unbiased molecular simulation data, an unsupervised deep learning method measures the differences in protein dynamics at a ligand-binding site depending on the bound ligands. A dimension reduction method extracts a dynamic feature that strongly correlates to the binding affinities. Moreover, the residues that play important roles in protein–ligand interactions are specified based on their contribution to the differences. These results indicate the potential for binding dynamics-based drug discovery.

Differences in ligand-induced protein dynamics extracted as a single feature from a deep learning-based analysis of MD simulations correlate with ligand binding affinity.

Delta Machine Learning to Improve Scoring-Ranking-Screening Performances of Protein-Ligand Scoring Functions

Protein-ligand scoring functions are widely used in structure-based drug design for fast evaluation of protein-ligand interactions, and it is of strong interest to develop scoring functions with machine-learning approaches. In this work, by expanding the training set, developing physically meaningful features, employing our recently developed linear empirical scoring function Lin_F9 (Yang, C. J. Chem. Inf. Model. 2021, 61, 4630-4644) as the baseline, and applying extreme gradient boosting (XGBoost) with Δ-machine learning, we have further improved the robustness and applicability of machine-learning scoring functions. Besides the top performances for scoring-ranking-screening power tests of the CASF-2016 benchmark, the new scoring function Δ_{Lin_F9}XGB also achieves superior scoring and ranking performances in different structure types that mimic real docking applications. The scoring powers of Δ_{Lin_F9}XGB for locally optimized poses, flexible redocked poses, and ensemble docked poses of the CASF-2016 core set achieve Pearson's correlation coefficient (R) values of 0.853, 0.839, and 0.813, respectively. In addition, the large-scale docking-based virtual screening test on the LIT-PCBA data set demonstrates the reliability and robustness of Δ_{Lin_F9}XGB in virtual screening application. The Δ_{Lin_F9}XGB scoring function and its code are freely available on the web at (https://yzhang.hpc.nyu.edu/Delta_LinF9_XGB).

Omicron BA.2 (B.1.1.529.2): High Potential for Becoming the Next Dominant Variant

The Omicron variant has three subvariants: BA.1 (B.1.1.529.1), BA.2 (B.1.1.529.2), and BA.3 (B.1.1.529.3). BA.2 is found to be able to alarmingly reinfect patients originally infected by Omicron BA.1. An important question is whether BA.2 or BA.3 will become a new dominating "variant of concern". Currently, no experimental data has been reported about BA.2 and BA.3. We construct a novel algebraic topology-based deep learning model to systematically evaluate BA.2's and BA.3's infectivity, vaccine breakthrough capability, and antibody resistance. Our comparative analysis of all main variants, namely, Alpha, Beta, Gamma, Delta, Lambda, Mu, BA.1, BA.2, and BA.3, unveils that BA.2 is about 1.5 and 4.2 times as contagious as BA.1 and Delta, respectively. It is also 30% and 17-fold more capable than BA.1 and Delta, respectively, to escape current vaccines. Therefore, we project that Omicron BA.2 is on a path to becoming the next dominant variant. We forecast that like Omicron BA.1, BA.2 will also seriously compromise most existing monoclonal antibodies. All key predictions have been nearly perfectly confirmed before the official publication of this work.

A simple spatial extension to the extended connectivity interaction features for binding affinity prediction

The representation of the protein-ligand complexes used in building machine learning models play an important role in the accuracy of binding affinity prediction. The Extended Connectivity Interaction Features (ECIF) is one such representation. We report that (i) including the discretized distances between protein-ligand atom pairs in the ECIF scheme improves predictive accuracy, and (ii) in an evaluation using gradient boosted trees, we found that the resampling method used in selecting the best hyperparameters has a strong effect on predictive performance, especially for benchmarking purposes.

Mathematical artificial intelligence design of mutation-proof COVID-19 monoclonal antibodies

Emerging severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants have compromised existing vaccines and posed a grand challenge to coronavirus disease 2019 (COVID-19) prevention, control, and global economic recovery. For COVID-19 patients, one of the most effective COVID-19 medications is monoclonal antibody (mAb) therapies. The United States Food and Drug Administration (U.S. FDA) has given the emergency use authorization (EUA) to a few mAbs, including those from Regeneron, Eli Elly, etc. However, they are also undermined by SARS-CoV-2 mutations. It is imperative to develop effective mutation-proof mAbs for treating COVID-19 patients infected by all emerging variants and/or the original SARS-CoV-2. We carry out a deep mutational scanning to present the blueprint of such mAbs using algebraic topology and artificial intelligence (AI). To reduce the risk of clinical trial-related failure, we select five mAbs either with FDA EUA or in clinical trials as our starting point. We demonstrate that topological AI-designed mAbs are effective to variants of concerns and variants of interest designated by the World Health Organization (WHO), as well as the original SARS-CoV-2. Our topological AI methodologies have been validated by tens of thousands of deep mutational data and their predictions have been confirmed by results from tens of experimental laboratories and population-level statistics of genome isolates from hundreds of thousands of patients.

Dowker complex based machine learning (DCML) models for protein-ligand binding affinity prediction

With the great advancements in experimental data, computational power and learning algorithms, artificial intelligence (AI) based drug design has begun to gain momentum recently. AI-based drug design has great promise to revolutionize pharmaceutical industries by significantly reducing the time and cost in drug discovery processes. However, a major issue remains for all AI-based learning model that is efficient molecular representations. Here we propose Dowker complex (DC) based molecular interaction representations and Riemann Zeta function based molecular featurization, for the first time. Molecular interactions between proteins and ligands (or others) are modeled as Dowker complexes. A multiscale representation is generated by using a filtration process, during which a series of DCs are generated at different scales. Combinatorial (Hodge) Laplacian matrices are constructed from these DCs, and the Riemann zeta functions from their spectral information can be used as molecular descriptors. To validate our models, we consider protein-ligand binding affinity prediction. Our DC-based machine learning (DCML) models, in particular, DC-based gradient boosting tree (DC-GBT), are tested on three most-commonly used datasets, i.e., including PDBbind-2007, PDBbind-2013 and PDBbind-2016, and extensively compared with other existing state-of-the-art models. It has been found that our DC-based descriptors can achieve the state-of-the-art results and have better performance than all machine learning models with traditional molecular descriptors. Our Dowker complex based machine learning models can be used in other tasks in AI-based drug design and molecular data analysis.

With the ever-increasing accumulation of chemical and biomolecular data, data-driven artificial intelligence (AI) models will usher in an era of faster, cheaper and more-efficient drug design and drug discovery. However, unlike image, text, video, audio data, molecular data from chemistry and biology, have much complicated three-dimensional structures, as well as physical and chemical properties. Efficient molecular representations and descriptors are key to the success of machine learning models in drug design. Here, we propose Dowker complex based molecular representation and Riemann Zeta function based molecular featurization, for the first time. To characterize the complicated molecular structures and interactions at the atomic level, Dowker complexes are constructed. Based on them, intrinsic mathematical invariants are derived and used as molecular descriptors, which can be further combined with machine learning and deep learning models. Our model has achieved state-of-the-art results in protein-ligand binding affinity prediction, demonstrating its great potential for other drug design and discovery problems.

Collaborative Approach between Explainable Artificial Intelligence and Simplified Chemical Interactions to Explore Active Ligands for Cyclin-Dependent Kinase 2

To improve virtual screening for drug discovery, we present a collaborative approach between explainable artificial intelligence (AI) and simplified chemical interaction scores to efficiently search for active ligands bound to the target receptor. In particular, we focus on cyclin-dependent kinase 2 (CDK2), which is well known as a cancer target protein. Docking simulation alone is insufficient to distinguish active ligands from decoy molecules. To identify active ligands, in this paper, machine learning is employed together with scoring functions that simplify the screened Coulomb and Lennard-Jones interactions between the ligands and residues of the target receptor. We demonstrate that these simplified interaction scores can significantly improve the classification ability of machine learning models. We also demonstrate that explainable AI together with the simplified scoring method can highlight the important residues of CDK2 for recognizing active ligands.

NeuralDock: Rapid and Conformation-Agnostic Docking of Small Molecules

Virtual screening is a cost- and time-effective alternative to traditional high-throughput screening in the drug discovery process. Both virtual screening approaches, structure-based molecular docking and ligand-based cheminformatics, suffer from computational cost, low accuracy, and/or reliance on prior knowledge of a ligand that binds to a given target. Here, we propose a neural network framework, NeuralDock, which accelerates the process of high-quality computational docking by a factor of 106, and does not require prior knowledge of a ligand that binds to a given target. By approximating both protein-small molecule conformational sampling and energy-based scoring, NeuralDock accurately predicts the binding energy, and affinity of a protein-small molecule pair, based on protein pocket 3D structure and small molecule topology. We use NeuralDock and 25 GPUs to dock 937 million molecules from the ZINC database against superoxide dismutase-1 in 21 h, which we validate with physical docking using MedusaDock. Due to its speed and accuracy, NeuralDock may be useful in brute-force virtual screening of massive chemical libraries and training of generative drug models.

Structure-Aware Multimodal Deep Learning for Drug-Protein Interaction Prediction

Identifying drug-protein interactions (DPIs) is crucial in drug discovery, and a number of machine learning methods have been developed to predict DPIs. Existing methods usually use unrealistic data sets with hidden bias, which will limit the accuracy of virtual screening methods. Meanwhile, most DPI prediction methods pay more attention to molecular representation but lack effective research on protein representation and high-level associations between different instances. To this end, we present the novel structure-aware multimodal deep DPI prediction model, STAMP-DPI, which was trained on a curated industry-scale benchmark data set. We built a high-quality benchmark data set named GalaxyDB for DPI prediction. This industry-scale data set along with an unbiased training procedure resulted in a more robust benchmark study. For informative protein representation, we constructed a structure-aware graph neural network method from the protein sequence by combining predicted contact maps and graph neural networks. Through further integration of structure-based representation and high-level pretrained embeddings for molecules and proteins, our model effectively captures the feature representation of the interactions between them. As a result, STAMP-DPI outperformed state-of-the-art DPI prediction methods by decreasing 7.00% mean square error (MSE) in the Davis data set and improving 8.89% area under the curve (AUC) in the GalaxyDB data set. Moreover, our model is an interpretable model with the transformer-based interaction mechanism, which can accurately reveal the binding sites between molecules and proteins.

Omicron BA.2 (B.1.1.529.2): high potential to becoming the next dominating variant

The Omicron variant of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has rapidly replaced the Delta variant as a dominating SARS-CoV-2 variant because of natural selection, which favors the variant with higher infectivity and stronger vaccine breakthrough ability. Omicron has three lineages or subvariants, BA.1 (B.1.1.529.1), BA.2 (B.1.1.529.2), and BA.3 (B.1.1.529.3). Among them, BA.1 is the currently prevailing subvariant. BA.2 shares 32 mutations with BA.1 but has 28 distinct ones. BA.3 shares most of its mutations with BA.1 and BA.2 except for one. BA.2 is found to be able to alarmingly reinfect patients originally infected by Omicron BA.1. An important question is whether BA.2 or BA.3 will become a new dominating ``variant of concern''. Currently, no experimental data has been reported about BA.2 and BA.3. We construct a novel algebraic topology-based deep learning model trained with tens of thousands of mutational and deep mutational data to systematically evaluate BA.2's and BA.3's infectivity, vaccine breakthrough capability, and antibody resistance. Our comparative analysis of all main variants namely, Alpha, Beta, Gamma, Delta, Lambda, Mu, BA.1, BA.2, and BA.3, unveils that BA.2 is about 1.5 and 4.2 times as contagious as BA.1 and Delta, respectively. It is also 30% and 17-fold more capable than BA.1 and Delta, respectively, to escape current vaccines. Therefore, we project that Omicron BA.2 is on its path to becoming the next dominating variant. We forecast that like Omicron BA.1, BA.2 will also seriously compromise most existing mAbs, except for sotrovimab developed by GlaxoSmithKline.

Persistent spectral based ensemble learning (PerSpect-EL) for protein-protein binding affinity prediction

Protein-protein interactions (PPIs) play a significant role in nearly all cellular and biological activities. Data-driven machine learning models have demonstrated great power in PPIs. However, the design of efficient molecular featurization poses a great challenge for all learning models for PPIs. Here, we propose persistent spectral (PerSpect) based PPI representation and featurization, and PerSpect-based ensemble learning (PerSpect-EL) models for PPI binding affinity prediction, for the first time. In our model, a sequence of Hodge (or combinatorial) Laplacian (HL) matrices at various different scales are generated from a specially designed filtration process. PerSpect attributes, which are statistical and combinatorial properties of spectrum information from these HL matrices, are used as features for PPI characterization. Each PerSpect attribute is input into a 1D convolutional neural network (CNN), and these CNN networks are stacked together in our PerSpect-based ensemble learning models. We systematically test our model on the two most commonly used datasets, i.e. SKEMPI and AB-Bind. It has been found that our model can achieve state-of-the-art results and outperform all existing models to the best of our knowledge.

Yuel: Improving the Generalizability of Structure-Free Compound-Protein Interaction Prediction

Predicting binding affinities between small molecules and the protein target is at the core of computational drug screening and drug target identification. Deep learning-based approaches have recently been adapted to predict binding affinities and they claim to achieve high prediction accuracy in their tests; we show that these approaches do not generalize, that is, they fail to predict interactions between unknown proteins and unknown small molecules. To address these shortcomings, we develop a new compound-protein interaction predictor, Yuel, which predicts compound-protein interactions with a higher generalizability than the existing methods. Upon comprehensive tests on various data sets, we find that out of all the deep-learning approaches surveyed, Yuel manifests the best ability to predict interactions between unknown compounds and unknown proteins.

Omicron BA.2 (B.1.1.529.2): high potential to becoming the next dominating variant

The Omicron variant of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has rapidly replaced the Delta variant as a dominating SARS-CoV-2 variant because of natural selection, which favors the variant with higher infectivity and stronger vaccine breakthrough ability. Omicron has three lineages or subvariants, BA.1 (B.1.1.529.1), BA.2 (B.1.1.529.2), and BA.3 (B.1.1.529.3). Among them, BA.1 is the currently prevailing subvariant. BA.2 shares 32 mutations with BA.1 but has 28 distinct ones. BA.3 shares most of its mutations with BA.1 and BA.2 except for one. BA.2 is found to be able to alarmingly reinfect patients originally infected by Omicron BA.1. An important question is whether BA.2 or BA.3 will become a new dominating "variant of concern". Currently, no experimental data has been reported about BA.2 and BA.3. We construct a novel algebraic topology-based deep learning model trained with tens of thousands of mutational and deep mutational data to systematically evaluate BA.2's and BA.3's infectivity, vaccine breakthrough capability, and antibody resistance. Our comparative analysis of all main variants namely, Alpha, Beta, Gamma, Delta, Lambda, Mu, BA.1, BA.2, and BA.3, unveils that BA.2 is about 1.5 and 4.2 times as contagious as BA.1 and Delta, respectively. It is also 30% and 17-fold more capable than BA.1 and Delta, respectively, to escape current vaccines. Therefore, we project that Omicron BA.2 is on its path to becoming the next dominating variant. We forecast that like Omicron BA.1, BA.2 will also seriously compromise most existing mAbs, except for sotrovimab developed by GlaxoSmithKline.

Molecular persistent spectral image (Mol-PSI) representation for machine learning models in drug design

Artificial intelligence (AI)-based drug design has great promise to fundamentally change the landscape of the pharmaceutical industry. Even though there are great progress from handcrafted feature-based machine learning models, 3D convolutional neural networks (CNNs) and graph neural networks, effective and efficient representations that characterize the structural, physical, chemical and biological properties of molecular structures and interactions remain to be a great challenge. Here, we propose an equal-sized molecular 2D image representation, known as the molecular persistent spectral image (Mol-PSI), and combine it with CNN model for AI-based drug design. Mol-PSI provides a unique one-to-one image representation for molecular structures and interactions. In general, deep models are empowered to achieve better performance with systematically organized representations in image format. A well-designed parallel CNN architecture for adapting Mol-PSIs is developed for protein-ligand binding affinity prediction. Our results, for the three most commonly used databases, including PDBbind-v2007, PDBbind-v2013 and PDBbind-v2016, are better than all traditional machine learning models, as far as we know. Our Mol-PSI model provides a powerful molecular representation that can be widely used in AI-based drug design and molecular data analysis.

Machine learning-based prediction of drug and ligand binding in BCL-2 variants through molecular dynamics

Venetoclax is a BH3 (BCL-2 Homology 3) mimetic used to treat leukemia and lymphoma by inhibiting the anti-apoptotic BCL-2 protein thereby promoting apoptosis of cancerous cells. Acquired resistance to Venetoclax via specific variants in BCL-2 is a major problem for the successful treatment of cancer patients. Replica exchange molecular dynamics (REMD) simulations combined with machine learning were used to define the average structure of variants in aqueous solution to predict changes in drug and ligand binding in BCL-2 variants. The variant structures all show shifts in residue positions that occlude the binding groove, and these are the primary contributors to drug resistance. Correspondingly, we established a method that can predict the severity of a variant as measured by the inhibitory constant (Ki) of Venetoclax by measuring the structure deviations to the binding cleft. In addition, we also applied machine learning to the phi and psi angles of the amino acid backbone to the ensemble of conformations that demonstrated a generalizable method for drug resistant predictions of BCL-2 proteins that elucidates changes where detailed understanding of the structure-function relationship is less clear.

MLIMC: Machine Learning-Based Implicit-Solvent Monte Carlo

Monte Carlo (MC) methods are important computational tools for molecular structure optimizations and predictions. When solvent effects are explicitly considered, MC methods become very expensive due to the large degree of freedom associated with the water molecules and mobile ions. Alternatively implicit-solvent MC can largely reduce the computational cost by applying a mean field approximation to solvent effects and meanwhile maintains the atomic detail of the target molecule. The two most popular implicit-solvent models are the Poisson-Boltzmann (PB) model and the Generalized Born (GB) model in a way such that the GB model is an approximation to the PB model but is much faster in simulation time. In this work, we develop a machine learning-based implicit-solvent Monte Carlo (MLIMC) method by combining the advantages of both implicit solvent models in accuracy and efficiency. Specifically, the MLIMC method uses a fast and accurate PB-based machine learning (PBML) scheme to compute the electrostatic solvation free energy at each step. We validate our MLIMC method by using a benzene-water system and a protein-water system. We show that the proposed MLIMC method has great advantages in speed and accuracy for molecular structure optimization and prediction.

InteractionGraphNet: A Novel and Efficient Deep Graph Representation Learning Framework for Accurate Protein-Ligand Interaction Predictions

Accurate quantification of protein-ligand interactions remains a key challenge to structure-based drug design. However, traditional machine learning (ML)-based methods based on handcrafted descriptors, one-dimensional protein sequences, and/or two-dimensional graph representations limit their capability to learn the generalized molecular interactions in 3D space. Here, we proposed a novel deep graph representation learning framework named InteractionGraphNet (IGN) to learn the protein-ligand interactions from the 3D structures of protein-ligand complexes. In IGN, two independent graph convolution modules were stacked to sequentially learn the intramolecular and intermolecular interactions, and the learned intermolecular interactions can be efficiently used for subsequent tasks. Extensive binding affinity prediction, large-scale structure-based virtual screening, and pose prediction experiments demonstrated that IGN achieved better or competitive performance against other state-of-the-art ML-based baselines and docking programs. More importantly, such state-of-the-art performance was proven from the successful learning of the key features in protein-ligand interactions instead of just memorizing certain biased patterns from data.

Prediction of Binding Free Energy of Protein-Ligand Complexes with a Hybrid Molecular Mechanics/Generalized Born Surface Area and Machine Learning Method

Accurate prediction of protein-ligand binding free energies is important in enzyme engineering and drug discovery. The molecular mechanics/generalized Born surface area (MM/GBSA) approach is widely used to estimate ligand-binding affinities, but its performance heavily relies on the accuracy of its energy components. A hybrid strategy combining MM/GBSA and machine learning (ML) has been developed to predict the binding free energies of protein-ligand systems. Based on the MM/GBSA energy terms and several features associated with protein-ligand interactions, our ML-based scoring function, GXLE, shows much better performance than MM/GBSA without entropy. In particular, the good transferability of the GXLE model is highlighted by its good performance in ranking power for prediction of the binding affinity of different ligands for either the docked structures or crystal structures. The GXLE scoring function and its code are freely available and can be used to correct the binding free energies computed by MM/GBSA.

LCP: Simple Representation of Docking Poses for Machine Learning: A Case Study on Xanthine Oxidase Inhibitors

In this paper, we propose a simple descriptor called the ligand coordinate profile (LCP) for describing docking poses. The LCP descriptor is generated from the coordinates of the polar hydrogen and heavy atoms of the docked ligand. We hypothesize that the prediction of binding poses can be enhanced through the combination of machine learning methods with the LCP descriptor. Two docking programs were used to predict ligand docking against xanthine oxidase. Four machine learning methods-k-nearest neighbors, random forest, support vector machine, and LightGBM-were used to determine whether machine learning-based models could be used to accurately identify the correct binding poses. Regardless of the machine learning method employed, the LCP descriptor demonstrated improved performance compared to the existing descriptor. The results of the leave-one-pdb-out approach revealed that the influence of the pose descriptor was also significant, as demonstrated through cross-validation. When evaluated using top-N metrics, the machine learning models were generally more effective than the docking programs. In addition, the LCP-based models outperformed those based on the existing descriptor. The results obtained in this study suggest that our proposed binding pose descriptor is effective for improving the docking accuracy of xanthine oxidase inhibitors.

A point cloud-based deep learning strategy for protein-ligand binding affinity prediction

There is great interest to develop artificial intelligence-based protein-ligand binding affinity models due to their immense applications in drug discovery. In this paper, PointNet and PointTransformer, two pointwise multi-layer perceptrons have been applied for protein-ligand binding affinity prediction for the first time. Three-dimensional point clouds could be rapidly generated from PDBbind-2016 with 3772 and 11 327 individual point clouds derived from the refined or/and general sets, respectively. These point clouds (the refined or the extended set) were used to train PointNet or PointTransformer, resulting in protein-ligand binding affinity prediction models with Pearson correlation coefficients R = 0.795 or 0.833 from the extended data set, respectively, based on the CASF-2016 benchmark test. The analysis of parameters suggests that the two deep learning models were capable to learn many interactions between proteins and their ligands, and some key atoms for the interactions could be visualized. The protein-ligand interaction features learned by PointTransformer could be further adapted for the XGBoost-based machine learning algorithm, resulting in prediction models with an average Rp of 0.827, which is on par with state-of-the-art machine learning models. These results suggest that the point clouds derived from PDBbind data sets are useful to evaluate the performance of 3D point clouds-centered deep learning algorithms, which could learn atomic features of protein-ligand interactions from natural evolution or medicinal chemistry and thus have wide applications in chemistry and biology.

Improving Structure-Based Virtual Screening with Ensemble Docking and Machine Learning

One of the main challenges of structure-based virtual screening (SBVS) is the incorporation of the receptor's flexibility, as its explicit representation in every docking run implies a high computational cost. Therefore, a common alternative to include the receptor's flexibility is the approach known as ensemble docking. Ensemble docking consists of using a set of receptor conformations and performing the docking assays over each of them. However, there is still no agreement on how to combine the ensemble docking results to obtain the final ligand ranking. A common choice is to use consensus strategies to aggregate the ensemble docking scores, but these strategies exhibit slight improvement regarding the single-structure approach. Here, we claim that using machine learning (ML) methodologies over the ensemble docking results could improve the predictive power of SBVS. To test this hypothesis, four proteins were selected as study cases: CDK2, FXa, EGFR, and HSP90. Protein conformational ensembles were built from crystallographic structures, whereas the evaluated compound library comprised up to three benchmarking data sets (DUD, DEKOIS 2.0, and CSAR-2012) and cocrystallized molecules. Ensemble docking results were processed through 30 repetitions of 4-fold cross-validation to train and validate two ML classifiers: logistic regression and gradient boosting trees. Our results indicate that the ML classifiers significantly outperform traditional consensus strategies and even the best performance case achieved with single-structure docking. We provide statistical evidence that supports the effectiveness of ML to improve the ensemble docking performance.

RealVS: Toward Enhancing the Precision of Top Hits in Ligand-Based Virtual Screening of Drug Leads from Large Compound Databases

Accurate modeling of compound bioactivities is essential for the virtual screening of drug leads. In real-world scenarios, pharmacists tend to choose from the top-k hit compounds ranked by predicted bioactivities from a large database with interest to continue wet experiments for drug discovery. Significant improvement of the precision of the top hits in ligand-based virtual screening of drug leads is more valuable than conventional schemes for accurately predicting the bioactivities of all compounds from a large database. Here, we proposed a new method, RealVS, to significantly improve the top hits' precision and learn interpretable key substructures associated with compound bioactivities. The features of RealVS involve the following points. (1) Abundant transferable information from the source domain was introduced for alleviating the insufficiency of inactive ligands associated with drug targets. (2) The adversarial domain alignment was adopted to fit the distribution of generated features of compounds from the training data set and that from the screening database for greater model generalization ability. (3) A novel objective function was proposed to simultaneously optimize the classification loss, regression loss, and adversarial loss, where most inactive ligands tend to be screened out before activity regression prediction. (4) Graph attention networks were adopted for learning key substructures associated with ligand bioactivities for better model interpretability. The results on a large number of benchmark data sets show that our method has significantly improved the precision of top hits under various k values in ligand-based virtual screening of drug leads from large compound databases, which is of great value in real-world scenarios. The web server of RealVS is freely available at noveldelta.com/RealVS for academic purposes, where virtual screening of hits from large compound databases is accessible.

EVOLUTIONARY DE RHAM-HODGE METHOD

The de Rham-Hodge theory is a landmark of the 20^th Century's mathematics and has had a great impact on mathematics, physics, computer science, and engineering. This work introduces an evolutionary de Rham-Hodge method to provide a unified paradigm for the multiscale geometric and topological analysis of evolving manifolds constructed from a filtration, which induces a family of evolutionary de Rham complexes. While the present method can be easily applied to close manifolds, the emphasis is given to more challenging compact manifolds with 2-manifold boundaries, which require appropriate analysis and treatment of boundary conditions on differential forms to maintain proper topological properties. Three sets of unique evolutionary Hodge Laplacians are proposed to generate three sets of topology-preserving singular spectra, for which the multiplicities of zero eigenvalues correspond to exactly the persistent Betti numbers of dimensions 0, 1 and 2. Additionally, three sets of non-zero eigenvalues further reveal both topological persistence and geometric progression during the manifold evolution. Extensive numerical experiments are carried out via the discrete exterior calculus to demonstrate the potential of the proposed paradigm for data representation and shape analysis of both point cloud data and density maps. To demonstrate the utility of the proposed method, the application is considered to the protein B-factor predictions of a few challenging cases for which existing biophysical models break down.

Learning protein-ligand binding affinity with atomic environment vectors

Scoring functions for the prediction of protein-ligand binding affinity have seen renewed interest in recent years when novel machine learning and deep learning methods started to consistently outperform classical scoring functions. Here we explore the use of atomic environment vectors (AEVs) and feed-forward neural networks, the building blocks of several neural network potentials, for the prediction of protein-ligand binding affinity. The AEV-based scoring function, which we term AEScore, is shown to perform as well or better than other state-of-the-art scoring functions on binding affinity prediction, with an RMSE of 1.22 pK units and a Pearson’s correlation coefficient of 0.83 for the CASF-2016 benchmark. However, AEScore does not perform as well in docking and virtual screening tasks, for which it has not been explicitly trained. Therefore, we show that the model can be combined with the classical scoring function AutoDock Vina in the context of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta$$\end{document}Δ-learning, where corrections to the AutoDock Vina scoring function are learned instead of the protein-ligand binding affinity itself. Combined with AutoDock Vina, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta$$\end{document}Δ-AEScore has an RMSE of 1.32 pK units and a Pearson’s correlation coefficient of 0.80 on the CASF-2016 benchmark, while retaining the docking and screening power of the underlying classical scoring function.

DEELIG: A Deep Learning Approach to Predict Protein-Ligand Binding Affinity

Protein-ligand binding prediction has extensive biological significance. Binding affinity helps in understanding the degree of protein-ligand interactions and is a useful measure in drug design. Protein-ligand docking using virtual screening and molecular dynamic simulations are required to predict the binding affinity of a ligand to its cognate receptor. Performing such analyses to cover the entire chemical space of small molecules requires intense computational power. Recent developments using deep learning have enabled us to make sense of massive amounts of complex data sets where the ability of the model to “learn” intrinsic patterns in a complex plane of data is the strength of the approach. Here, we have incorporated convolutional neural networks to find spatial relationships among data to help us predict affinity of binding of proteins in whole superfamilies toward a diverse set of ligands without the need of a docked pose or complex as user input. The models were trained and validated using a stringent methodology for feature extraction. Our model performs better in comparison to some existing methods used widely and is suitable for predictions on high-resolution protein crystal (⩽2.5 Å) and nonpeptide ligand as individual inputs. Our approach to network construction and training on protein-ligand data set prepared in-house has yielded significant insights. We have also tested DEELIG on few COVID-19 main protease-inhibitor complexes relevant to the current public health scenario. DEELIG-based predictions can be incorporated in existing databases including RSCB PDB, PDBMoad, and PDBbind in filling missing binding affinity data for protein-ligand complexes.

Extended connectivity interaction features: improving binding affinity prediction through chemical description

MOTIVATION

Machine-learning scoring functions (SFs) have been found to outperform standard SFs for binding affinity prediction of protein-ligand complexes. A plethora of reports focus on the implementation of increasingly complex algorithms, while the chemical description of the system has not been fully exploited.

RESULTS

Herein, we introduce Extended Connectivity Interaction Features (ECIF) to describe protein-ligand complexes and build machine-learning SFs with improved predictions of binding affinity. ECIF are a set of protein-ligand atom-type pair counts that take into account each atom's connectivity to describe it and thus define the pair types. ECIF were used to build different machine-learning models to predict protein-ligand affinities (pKd/pKi). The models were evaluated in terms of 'scoring power' on the Comparative Assessment of Scoring Functions 2016. The best models built on ECIF achieved Pearson correlation coefficients of 0.857 when used on its own, and 0.866 when used in combination with ligand descriptors, demonstrating ECIF descriptive power.

AVAILABILITY AND IMPLEMENTATION

Data and code to reproduce all the results are freely available at https://github.com/DIFACQUIM/ECIF.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Computational representations of protein-ligand interfaces for structure-based virtual screening

Introduction: Structure-based virtual screening (SBVS) is an essential strategy for hit identification. SBVS primarily uses molecular docking, which exploits the protein-ligand binding mode and associated affinity score for compound ranking. Previous studies have shown that computational representation of protein-ligand interfaces and the later establishment of machine learning models are efficacious in improving the accuracy of SBVS.Areas covered: The authors review the computational methods for representing protein-ligand interfaces, which include the traditional ones that use deliberately designed fingerprints and descriptors and the more recent methods that automatically extract features with deep learning. The effects of these methods on the performance of machine learning models are briefly discussed. Additionally, case studies that applied various computational representations to machine learning are cited with remarks.Expert opinion: It has become a trend to extract binding features automatically by deep learning, which uses a completely end-to-end representation. However, there is still plenty of scope for improvement . The interpretability of deep-learning models, the organization of data management, the quantity and quality of available data, and the optimization of hyperparameters could impact the accuracy of feature extraction. In addition, other important structural factors such as water molecules and protein flexibility should be considered.

DeepBSP-a Machine Learning Method for Accurate Prediction of Protein-Ligand Docking Structures

In recent years, machine-learning-based scoring functions have significantly improved the scoring power. However, many of these methods do not perform well in distinguishing the native structure from docked decoy poses due to the lack of decoy structural information in their training data. Here, we developed a machine-learning model, named DeepBSP, that can directly predict the root mean square deviation (rmsd) of a ligand docking pose with reference to its native binding pose. Unlike the binding affinity, the rmsd between the docking poses with reference to their native structures can be straightforwardly determined. By training on a generated data set with 11,925 native complexes and more than 165,000 docked poses, our model shows excellent docking power on our test set and also on the CASF-2016 docking decoy set compared to other major scoring functions. Thus, by combining molecular dockings that generate many poses with the application of DeepBSP, one can more accurately predict the best binding pose that is closest to the native complex structure. This DeepBSP model shall be very useful in picking out poses close to their natives from many poses generated from a dock application.

Review of COVID-19 Antibody Therapies

In the global health emergency caused by coronavirus disease 2019 (COVID-19), efficient and specific therapies are urgently needed. Compared with traditional small-molecular drugs, antibody therapies are relatively easy to develop; they are as specific as vaccines in targeting severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2); and they have thus attracted much attention in the past few months. This article reviews seven existing antibodies for neutralizing SARS-CoV-2 with 3D structures deposited in the Protein Data Bank (PDB). Five 3D antibody structures associated with the SARS-CoV spike (S) protein are also evaluated for their potential in neutralizing SARS-CoV-2. The interactions of these antibodies with the S protein receptor-binding domain (RBD) are compared with those between angiotensin-converting enzyme 2 and RBD complexes. Due to the orders of magnitude in the discrepancies of experimental binding affinities, we introduce topological data analysis, a variety of network models, and deep learning to analyze the binding strength and therapeutic potential of the 14 antibody-antigen complexes. The current COVID-19 antibody clinical trials, which are not limited to the S protein target, are also reviewed.

Forman persistent Ricci curvature (FPRC)-based machine learning models for protein-ligand binding affinity prediction

Artificial intelligence (AI) techniques have already been gradually applied to the entire drug design process, from target discovery, lead discovery, lead optimization and preclinical development to the final three phases of clinical trials. Currently, one of the central challenges for AI-based drug design is molecular featurization, which is to identify or design appropriate molecular descriptors or fingerprints. Efficient and transferable molecular descriptors are key to the success of all AI-based drug design models. Here we propose Forman persistent Ricci curvature (FPRC)-based molecular featurization and feature engineering, for the first time. Molecular structures and interactions are modeled as simplicial complexes, which are generalization of graphs to their higher dimensional counterparts. Further, a multiscale representation is achieved through a filtration process, during which a series of nested simplicial complexes at different scales are generated. Forman Ricci curvatures (FRCs) are calculated on the series of simplicial complexes, and the persistence and variation of FRCs during the filtration process is defined as FPRC. Moreover, persistent attributes, which are FPRC-based functions and properties, are employed as molecular descriptors, and combined with machine learning models, in particular, gradient boosting tree (GBT). Our FPRC-GBT models are extensively trained and tested on three most commonly-used datasets, including PDBbind-2007, PDBbind-2013 and PDBbind-2016. It has been found that our results are better than the ones from machine learning models with traditional molecular descriptors.

Persistent spectral-based machine learning (PerSpect ML) for protein-ligand binding affinity prediction

Molecular descriptors are essential to not only quantitative structure-activity relationship (QSAR) models but also machine learning-based material, chemical, and biological data analysis. Here, we propose persistent spectral-based machine learning (PerSpect ML) models for drug design. Different from all previous spectral models, a filtration process is introduced to generate a sequence of spectral models at various different scales. PerSpect attributes are defined as the function of spectral variables over the filtration value. Molecular descriptors obtained from PerSpect attributes are combined with machine learning models for protein-ligand binding affinity prediction. Our results, for the three most commonly used databases including PDBbind-2007, PDBbind-2013, and PDBbind-2016, are better than all existing models, as far as we know. The proposed PerSpect theory provides a powerful feature engineering framework. PerSpect ML models demonstrate great potential to significantly improve the performance of learning models in molecular data analysis.

Computational Screening of Potential Inhibitors of Desulfobacter postgatei for Pyrite Scale Prevention in Oil and Gas Wells

Sulfate-reducing bacteria (SRB), such as Desulfobacter postgatei are found in oil wells. However, they lead to the release of hydrogen sulfide. This in turn leads to the iron sulfide scale formation (pyrite). ATP sulfurylase is an enzyme present in SRB, which catalyzes the formation of adenylyl sulfate (APS) and inorganic pyrophosphatase (PPi) from ATP and sulfate. This reaction is the first among many in hydrogen sulfide production by D. postgatei . Consensus scoring using molecular docking and machine learning was used to identify three potential inhibitors of ATP sulfurylase from a database of about 40 million compounds. These selected hits ((S,E)-1-(4-methoxyphenyl)-3-(9-((m-tolylimino)methyl)-9,10-dihydroanthracen-9-yl)pyrrolidine-2,5-dione; methyl 2-[[(1S)-5-cyano-2-imino-1-(4-phenylthiazol-2-yl)-3-azaspiro[5.5]undec-4-en-4-yl]sulfanyl]acetate; and (4S)-4-(3-chloro-4-hydroxy-phenyl)-1-(6-hydroxypyridazin-3-yl)-3-methyl-4,5-dihydropyrazolo[3,4-b]pyridin-6-ol), known as A, B, and C, respectively) all had good binding affinities with ATP sulfurylase and were further analyzed for their toxicological properties. Compound A had the highest docking score. However, based on the physicochemical and toxicological properties, only compound C was predicted to be both safe and effective as a potential inhibitor of ATP sulfurylase, hence the preferred choice. The molecular interactions of compound C revealed favorable interactions with the following residues: LEU213, ASP308, ARG307, TRP347, LEU224, GLN212, MET211, and HIS309.

Deep Learning in Virtual Screening: Recent Applications and Developments

Drug discovery is a cost and time-intensive process that is often assisted by computational methods, such as virtual screening, to speed up and guide the design of new compounds. For many years, machine learning methods have been successfully applied in the context of computer-aided drug discovery. Recently, thanks to the rise of novel technologies as well as the increasing amount of available chemical and bioactivity data, deep learning has gained a tremendous impact in rational active compound discovery. Herein, recent applications and developments of machine learning, with a focus on deep learning, in virtual screening for active compound design are reviewed. This includes introducing different compound and protein encodings, deep learning techniques as well as frequently used bioactivity and benchmark data sets for model training and testing. Finally, the present state-of-the-art, including the current challenges and emerging problems, are examined and discussed.

Persistent spectral hypergraph based machine learning (PSH-ML) for protein-ligand binding affinity prediction

Molecular descriptors are essential to not only quantitative structure activity/property relationship (QSAR/QSPR) models, but also machine learning based chemical and biological data analysis. In this paper, we propose persistent spectral hypergraph (PSH) based molecular descriptors or fingerprints for the first time. Our PSH-based molecular descriptors are used in the characterization of molecular structures and interactions, and further combined with machine learning models, in particular gradient boosting tree (GBT), for protein-ligand binding affinity prediction. Different from traditional molecular descriptors, which are usually based on molecular graph models, a hypergraph-based topological representation is proposed for protein-ligand interaction characterization. Moreover, a filtration process is introduced to generate a series of nested hypergraphs in different scales. For each of these hypergraphs, its eigen spectrum information can be obtained from the corresponding (Hodge) Laplacain matrix. PSH studies the persistence and variation of the eigen spectrum of the nested hypergraphs during the filtration process. Molecular descriptors or fingerprints can be generated from persistent attributes, which are statistical or combinatorial functions of PSH, and combined with machine learning models, in particular, GBT. We test our PSH-GBT model on three most commonly used datasets, including PDBbind-2007, PDBbind-2013 and PDBbind-2016. Our results, for all these databases, are better than all existing machine learning models with traditional molecular descriptors, as far as we know.

SMPLIP-Score: predicting ligand binding affinity from simple and interpretable on-the-fly interaction fingerprint pattern descriptors

In drug discovery, rapid and accurate prediction of protein–ligand binding affinities is a pivotal task for lead optimization with acceptable on-target potency as well as pharmacological efficacy. Furthermore, researchers hope for a high correlation between docking score and pose with key interactive residues, although scoring functions as free energy surrogates of protein–ligand complexes have failed to provide collinearity. Recently, various machine learning or deep learning methods have been proposed to overcome the drawbacks of scoring functions. Despite being highly accurate, their featurization process is complex and the meaning of the embedded features cannot directly be interpreted by human recognition without an additional feature analysis. Here, we propose SMPLIP-Score (Substructural Molecular and Protein–Ligand Interaction Pattern Score), a direct interpretable predictor of absolute binding affinity. Our simple featurization embeds the interaction fingerprint pattern on the ligand-binding site environment and molecular fragments of ligands into an input vectorized matrix for learning layers (random forest or deep neural network). Despite their less complex features than other state-of-the-art models, SMPLIP-Score achieved comparable performance, a Pearson’s correlation coefficient up to 0.80, and a root mean square error up to 1.18 in pK units with several benchmark datasets (PDBbind v.2015, Astex Diverse Set, CSAR NRC HiQ, FEP, PDBbind NMR, and CASF-2016). For this model, generality, predictive power, ranking power, and robustness were examined using direct interpretation of feature matrices for specific targets.

Ollivier Persistent Ricci Curvature-Based Machine Learning for the Protein-Ligand Binding Affinity Prediction

Efficient molecular featurization is one of the major issues for machine learning models in drug design. Here, we propose a persistent Ricci curvature (PRC), in particular, Ollivier PRC (OPRC), for the molecular featurization and feature engineering, for the first time. The filtration process proposed in the persistent homology is employed to generate a series of nested molecular graphs. Persistence and variation of Ollivier Ricci curvatures on these nested graphs are defined as OPRC. Moreover, persistent attributes, which are statistical and combinatorial properties of OPRCs during the filtration process, are used as molecular descriptors and further combined with machine learning models, in particular, gradient boosting tree (GBT). Our OPRC-GBT model is used in the prediction of the protein-ligand binding affinity, which is one of the key steps in drug design. Based on three of the most commonly used data sets from the well-established protein-ligand binding databank, that is, PDBbind, we intensively test our model and compare with existing models. It has been found that our model can achieve the state-of-the-art results and has advantages over traditional molecular descriptors.

Hypergraph-based persistent cohomology (HPC) for molecular representations in drug design

Artificial intelligence (AI) based drug design has demonstrated great potential to fundamentally change the pharmaceutical industries. Currently, a key issue in AI-based drug design is efficient transferable molecular descriptors or fingerprints. Here, we present hypergraph-based molecular topological representation, hypergraph-based (weighted) persistent cohomology (HPC/HWPC) and HPC/HWPC-based molecular fingerprints for machine learning models in drug design. Molecular structures and their atomic interactions are highly complicated and pose great challenges for efficient mathematical representations. We develop the first hypergraph-based topological framework to characterize detailed molecular structures and interactions at atomic level. Inspired by the elegant path complex model, hypergraph-based embedded homology and persistent homology have been proposed recently. Based on them, we construct HPC/HWPC, and use them to generate molecular descriptors for learning models in protein-ligand binding affinity prediction, one of the key step in drug design. Our models are tested on three most commonly-used databases, including PDBbind-v2007, PDBbind-v2013 and PDBbind-v2016, and outperform all existing machine learning models with traditional molecular descriptors. Our HPC/HWPC models have demonstrated great potential in AI-based drug design.

Site-Level Bioactivity of Small-Molecules from Deep-Learned Representations of Quantum Chemistry

Atom- or bond-level chemical properties of interest in medicinal chemistry, such as drug metabolism and electrophilic reactivity, are important to understand and predict across arbitrary new molecules. Deep learning can be used to map molecular structures to their chemical properties, but the data sets for these tasks are relatively small, which can limit accuracy and generalizability. To overcome this limitation, it would be preferable to model these properties on the basis of the underlying quantum chemical characteristics of small molecules. However, it is difficult to learn higher level chemical properties from lower level quantum calculations. To overcome this challenge, we pretrained deep learning models to compute quantum chemical properties and then reused the intermediate representations constructed by the pretrained network. Transfer learning, in this way, substantially outperformed models based on chemical graphs alone or quantum chemical properties alone. This result was robust, observable in five prediction tasks: identifying sites of epoxidation by metabolic enzymes and identifying sites of covalent reactivity with cyanide, glutathione, DNA and protein. We see that this approach may substantially improve the accuracy of deep learning models for specific chemical structures, such as aromatic systems.

Unveiling the molecular mechanism of SARS-CoV-2 main protease inhibition from 137 crystal structures using algebraic topology and deep learning

Currently, there is neither effective antiviral drugs nor vaccine for coronavirus disease 2019 (COVID-19) caused by acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Due to its high conservativeness and low similarity with human genes, SARS-CoV-2 main protease (Mpro) is one of the most favorable drug targets. However, the current understanding of the molecular mechanism of Mpro inhibition is limited by the lack of reliable binding affinity ranking and prediction of existing structures of Mpro-inhibitor complexes. This work integrates mathematics (i.e., algebraic topology) and deep learning (MathDL) to provide a reliable ranking of the binding affinities of 137 SARS-CoV-2 Mpro inhibitor structures. We reveal that Gly143 residue in Mpro is the most attractive site to form hydrogen bonds, followed by Glu166, Cys145, and His163. We also identify 71 targeted covalent bonding inhibitors. MathDL was validated on the PDBbind v2016 core set benchmark and a carefully curated SARS-CoV-2 inhibitor dataset to ensure the reliability of the present binding affinity prediction. The present binding affinity ranking, interaction analysis, and fragment decomposition offer a foundation for future drug discovery efforts.

Three-Dimensional Convolutional Neural Networks and a Cross-Docked Data Set for Structure-Based Drug Design

One of the main challenges in drug discovery is predicting protein-ligand binding affinity. Recently, machine learning approaches have made substantial progress on this task. However, current methods of model evaluation are overly optimistic in measuring generalization to new targets, and there does not exist a standard data set of sufficient size to compare performance between models. We present a new data set for structure-based machine learning, the CrossDocked2020 set, with 22.5 million poses of ligands docked into multiple similar binding pockets across the Protein Data Bank, and perform a comprehensive evaluation of grid-based convolutional neural network (CNN) models on this data set. We also demonstrate how the partitioning of the training data and test data can impact the results of models trained with the PDBbind data set, how performance improves by adding more lower-quality training data, and how training with docked poses imparts pose sensitivity to the predicted affinity of a complex. Our best performing model, an ensemble of five densely connected CNNs, achieves a root mean squared error of 1.42 and Pearson R of 0.612 on the affinity prediction task, an AUC of 0.956 at binding pose classification, and a 68.4% accuracy at pose selection on the CrossDocked2020 set. By providing data splits for clustered cross-validation and the raw data for the CrossDocked2020 set, we establish the first standardized data set for training machine learning models to recognize ligands in noncognate target structures while also greatly expanding the number of poses available for training. In order to facilitate community adoption of this data set for benchmarking protein-ligand binding affinity prediction, we provide our models, weights, and the CrossDocked2020 set at https://github.com/gnina/models.

Selecting machine-learning scoring functions for structure-based virtual screening

Interest in docking technologies has grown parallel to the ever increasing number and diversity of 3D models for macromolecular therapeutic targets. Structure-Based Virtual Screening (SBVS) aims at leveraging these experimental structures to discover the necessary starting points for the drug discovery process. It is now established that Machine Learning (ML) can strongly enhance the predictive accuracy of scoring functions for SBVS by exploiting large datasets from targets, molecules and their associations. However, with greater choice, the question of which ML-based scoring function is the most suitable for prospective use on a given target has gained importance. Here we analyse two approaches to select an existing scoring function for the target along with a third approach consisting in generating a scoring function tailored to the target. These analyses required discussing the limitations of popular SBVS benchmarks, the alternatives to benchmark scoring functions for SBVS and how to generate them or use them using freely-available software.

Mutations Strengthened SARS-CoV-2 Infectivity

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infectivity is a major concern in coronavirus disease 2019 (COVID-19) prevention and economic reopening. However, rigorous determination of SARS-CoV-2 infectivity is very difficult owing to its continuous evolution with over 10,000 single nucleotide polymorphisms (SNP) variants in many subtypes. We employ an algebraic topology-based machine learning model to quantitatively evaluate the binding free energy changes of SARS-CoV-2 spike glycoprotein (S protein) and host angiotensin-converting enzyme 2 receptor following mutations. We reveal that the SARS-CoV-2 virus becomes more infectious. Three out of six SARS-CoV-2 subtypes have become slightly more infectious, while the other three subtypes have significantly strengthened their infectivity. We also find that SARS-CoV-2 is slightly more infectious than SARS-CoV according to computed S protein-angiotensin-converting enzyme 2 binding free energy changes. Based on a systematic evaluation of all possible 3686 future mutations on the S protein receptor-binding domain, we show that most likely future mutations will make SARS-CoV-2 more infectious. Combining sequence alignment, probability analysis, and binding free energy calculation, we predict that a few residues on the receptor-binding motif, i.e., 452, 489, 500, 501, and 505, have high chances to mutate into significantly more infectious COVID-19 strains.