Binding affinity prediction largely determines the discovery efficiency of lead compounds in drug discovery. Recently, machine learning (ML)-based approaches have attracted much attention in hopes of enhancing the predictive performance of traditional physics-based approaches. In this study, we evaluated the impact of structural dynamic information on the binding affinity prediction by comparing the models trained on different dimensional descriptors, using three targets (i.e. JAK1, TAF1-BD2 and DDR1) and their corresponding ligands as the examples. Here, 2D descriptors are traditional ECFP4 fingerprints, 3D descriptors are the energy terms of the Smina and NNscore scoring functions and 4D descriptors contain the structural dynamic information derived from the trajectories based on molecular dynamics (MD) simulations. We systematically investigate the MD-refined binding affinity prediction performance of three classical ML algorithms (i.e. RF, SVR and XGB) as well as two common virtual screening methods, namely Glide docking and MM/PBSA. The outcomes of the ML models built using various dimensional descriptors and their combinations reveal that the MD refinement with the optimized protocol can improve the predictive performance on the TAF1-BD2 target with considerable structural flexibility, but not for the less flexible JAK1 and DDR1 targets, when taking docking poses as the initial structure instead of the crystal structures. The results highlight the importance of the initial structures to the final performance of the model through conformational analysis on the three targets with different flexibility.

Machine learning including modern deep learning models has been extensively used in drug design and screening. However, reliable prediction of molecular properties is still challenging when exploring out-of-domain regimes, even for deep neural networks. Therefore, it is important to understand the uncertainty of model predictions, especially when the predictions are used to guide further experiments. In this study, we explored the utility and effectiveness of evidential uncertainty in compound screening. The evidential Graphormer model was proposed for uncertainty-guided discovery of KDM1A/LSD1 inhibitors. The benchmarking results illustrated that (i) Graphormer exhibited comparative predictive power to state-of-the-art models, and (ii) evidential regression enabled well-ranked uncertainty estimates and calibrated predictions. Subsequently, we leveraged time-splitting on the curated KDM1A/LSD1 dataset to simulate out-of-distribution predictions. The retrospective virtual screening showed that the evidential uncertainties helped reduce false positives among the top-acquired compounds and thus enabled higher experimental validation rates. The trained model was then used to virtually screen an independent in-house compound set. The top 50 compounds ranked by two different ranking strategies were experimentally validated, respectively. In general, our study highlighted the importance to understand the uncertainty in prediction, which can be recognized as an interpretable dimension to model predictions.

Understanding drug selectivity mechanism is a long-standing issue for helping design drugs with high specificity. Designing drugs targeting cyclin-dependent kinases (CDKs) with high selectivity is challenging because of their highly conserved binding pockets. To reveal the underlying general selectivity mechanism, we carried out comprehensive analyses from both the thermodynamics and kinetics points of view on a representative CDK12 inhibitor. To fully capture the binding features of the drug-target recognition process, we proposed to use kinetic residue energy analysis (KREA) in conjunction with the community network analysis (CNA) to reveal the underlying cooperation effect between individual residues/protein motifs to the binding/dissociating process of the ligand. The general mechanism of drug selectivity in CDKs can be summarized as that the difference of structural cooperation between the ligand and the protein motifs leads to the difference of the energetic contribution of the key residues to the ligand. The proposed mechanisms may be prevalent in drug selectivity issues, and the insights may help design new strategies to overcome/attenuate the drug selectivity associated problems.

Drug design targeting protein-protein interactions (PPIs) associated with the development of diseases has been one of the most important therapeutic strategies. Besides interrupting the PPIs with PPI inhibitors/blockers, increasing evidence shows that stabilizing the interaction between two interacting proteins may also benefit the therapy, such as the development of various types of molecular glues/stabilizers that mostly work by stabilizing the two interacting proteins to regulate the downstream biological effects. However, characterizing the stabilization effect of a stabilizer is usually hard or too complicated for traditional experiments since it involves ternary interactions [protein-protein-stabilizer (PPS) interaction]. Thus, developing reliable computational strategies will facilitate the discovery/design of molecular glues or PPI stabilizers. Here, by fully analyzing the energetic features of the binary interactions in the PPS ternary complex, we systematically investigated the performance of molecular mechanics Poisson-Boltzmann surface area (MM/PBSA) and molecular mechanics generalized Born surface area (MM/GBSA) methods on characterizing the stabilization effects of stabilizers in 14-3-3 systems. The results show that both MM/PBSA and MM/GBSA are powerful tools in distinguishing the stabilizers from the decoys (with area under the curves of 0.90-0.93 for all tested cases) and are reasonable for ranking protein-peptide interactions in the presence or absence of stabilizers as well (with the average Pearson correlation coefficient of ~0.6 at a relatively high dielectric constant for both methods). Moreover, to give a detailed picture of the stabilization effects, the stabilization mechanism is also analyzed from the structural and energetic points of view for individual systems containing strong or weak stabilizers. This study demonstrates a potential strategy to accelerate the discovery of PPI stabilizers.

Predicting the native or near-native binding pose of a small molecule within a protein binding pocket is an extremely important task in structure-based drug design, especially in the hit-to-lead and lead optimization phases. In this study, fastDRH, a free and open accessed web server, was developed to predict and analyze protein-ligand complex structures. In fastDRH server, AutoDock Vina and AutoDock-GPU docking engines, structure-truncated MM/PB(GB)SA free energy calculation procedures and multiple poses based per-residue energy decomposition analysis were well integrated into a user-friendly and multifunctional online platform. Benefit from the modular architecture, users can flexibly use one or more of three features, including molecular docking, docking pose rescoring and hotspot residue prediction, to obtain the key information clearly based on a result analysis panel supported by 3Dmol.js and Apache ECharts. In terms of protein-ligand binding mode prediction, the integrated structure-truncated MM/PB(GB)SA rescoring procedures exhibit a success rate of >80% in benchmark, which is much better than the AutoDock Vina (~70%). For hotspot residue identification, our multiple poses based per-residue energy decomposition analysis strategy is a more reliable solution than the one using only a single pose, and the performance of our solution has been experimentally validated in several drug discovery projects. To summarize, the fastDRH server is a useful tool for predicting the ligand binding mode and the hotspot residue of protein for ligand binding. The fastDRH server is accessible free of charge at http://cadd.zju.edu.cn/fastdrh/.

Machine learning-based scoring functions (MLSFs) have attracted extensive attention recently and are expected to be potential rescoring tools for structure-based virtual screening (SBVS). However, a major concern nowadays is whether MLSFs trained for generic uses rather than a given target can consistently be applicable for VS. In this study, a systematic assessment was carried out to re-evaluate the effectiveness of 14 reported MLSFs in VS. Overall, most of these MLSFs could hardly achieve satisfactory results for any dataset, and they could even not outperform the baseline of classical SFs such as Glide SP. An exception was observed for RFscore-VS trained on the Directory of Useful Decoys-Enhanced dataset, which showed its superiority for most targets. However, in most cases, it clearly illustrated rather limited performance on the targets that were dissimilar to the proteins in the corresponding training sets. We also used the top three docking poses rather than the top one for rescoring and retrained the models with the updated versions of the training set, but only minor improvements were observed. Taken together, generic MLSFs may have poor generalization capabilities to be applicable for the real VS campaigns. Therefore, it should be quite cautious to use this type of methods for VS.

How to accurately estimate protein-ligand binding affinity remains a key challenge in computer-aided drug design (CADD). In many cases, it has been shown that the binding affinities predicted by classical scoring functions (SFs) cannot correlate well with experimentally measured biological activities. In the past few years, machine learning (ML)-based SFs have gradually emerged as potential alternatives and outperformed classical SFs in a series of studies. In this study, to better recognize the potential of classical SFs, we have conducted a comparative assessment of 25 commonly used SFs. Accordingly, the scoring power was systematically estimated by using the state-of-the-art ML methods that replaced the original multiple linear regression method to refit individual energy terms. The results show that the newly-developed ML-based SFs consistently performed better than classical ones. In particular, gradient boosting decision tree (GBDT) and random forest (RF) achieved the best predictions in most cases. The newly-developed ML-based SFs were also tested on another benchmark modified from PDBbind v2007, and the impacts of structural and sequence similarities were evaluated. The results indicated that the superiority of the ML-based SFs could be fully guaranteed when sufficient similar targets were contained in the training set. Moreover, the effect of the combinations of features from multiple SFs was explored, and the results indicated that combining NNscore2.0 with one to four other classical SFs could yield the best scoring power. However, it was not applicable to derive a generic target-specific SF or SF combination.

Although a wide variety of machine learning (ML) algorithms have been utilized to learn quantitative structure-activity relationships (QSARs), there is no agreed single best algorithm for QSAR learning. Therefore, a comprehensive understanding of the performance characteristics of popular ML algorithms used in QSAR learning is highly desirable. In this study, five linear algorithms [linear function Gaussian process regression (linear-GPR), linear function support vector machine (linear-SVM), partial least squares regression (PLSR), multiple linear regression (MLR) and principal component regression (PCR)], three analogizers [radial basis function support vector machine (rbf-SVM), K-nearest neighbor (KNN) and radial basis function Gaussian process regression (rbf-GPR)], six symbolists [extreme gradient boosting (XGBoost), Cubist, random forest (RF), multiple adaptive regression splines (MARS), gradient boosting machine (GBM), and classification and regression tree (CART)] and two connectionists [principal component analysis artificial neural network (pca-ANN) and deep neural network (DNN)] were employed to learn the regression-based QSAR models for 14 public data sets comprising nine physicochemical properties and five toxicity endpoints. The results show that rbf-SVM, rbf-GPR, XGBoost and DNN generally illustrate better performances than the other algorithms. The overall performances of different algorithms can be ranked from the best to the worst as follows: rbf-SVM > XGBoost > rbf-GPR > Cubist > GBM > DNN > RF > pca-ANN > MARS > linear-GPR ≈ KNN > linear-SVM ≈ PLSR > CART ≈ PCR ≈ MLR. In terms of prediction accuracy and computational efficiency, SVM and XGBoost are recommended to the regression learning for small data sets, and XGBoost is an excellent choice for large data sets. We then investigated the performances of the ensemble models by integrating the predictions of multiple ML algorithms. The results illustrate that the ensembles of two or three algorithms in different categories can indeed improve the predictions of the best individual ML algorithms.

SUMMARY: High-level quantum mechanics (QM) methods are no doubt the most reliable approaches for the prediction of atomic charges, but it usually needs very large computational resources, which apparently hinders the use of high-quality atomic charges in large-scale molecular modeling, such as high-throughput virtual screening. To solve this problem, several algorithms based on machine-learning (ML) have been developed to fit high-level QM atomic charges. Here, we proposed DeepChargePredictor, a web server that is able to generate the high-level QM atomic charges for small molecules based on two state-of-the-art ML algorithms developed in our group, namely AtomPathDescriptor and DeepAtomicCharge. These two algorithms were seamlessly integrated into the platform with the capability to predict three kinds of charges (i.e. RESP, AM1-BCC and DDEC) widely used in structure-based drug design. Moreover, we have comprehensively evaluated the performance of these charges generated by DeepChargePredictor for large-scale drug design applications, such as end-point binding free energy calculations and virtual screening, which all show reliable or even better performance compared with the baseline methods. AVAILABILITY AND IMPLEMENTATION: The data in the article can be obtained on the web page http://cadd.zju.edu.cn/deepchargepredictor/publication. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Protein kinases have been regarded as important therapeutic targets for many diseases. Currently, a total of 41 kinase inhibitors have been approved by the Food and Drug Administration, along with a large number of kinase inhibitors being evaluated in clinical and preclinical trials. Among all, allosteric inhibitors, such as type II kinase inhibitors, have attracted extensive attention owing to their potential high selectivity. Nowadays, molecular docking has become a powerful tool to search for novel kinase inhibitors. However, as for type II kinase inhibitors, their allosteric characteristics may exert a deep influence on docking accuracy. In this study, a comprehensive assessment was conducted to evaluate the effectiveness of nine docking algorithms towards type II kinase inhibitors. The calculation results showed that most tested docking programs, especially Glide with XP scoring, LeDock and Surflex-Dock, succeeded in the accurate identification of near-native binding poses, with the success rates ranging from 0.80 to 0.90, and the scoring functions in GOLD and LeDock outperformed the others in the prediction of relative binding affinities. In terms of the P-values, areas under the curve and enrichment factors, Glide with XP scoring, Surflex-Dock, GOLD with Astex Statistical Potential scoring and LeDock had better screening power to discriminate between active compounds and decoys. However, the screening power is sensitive to different initial conformations of the same target. It is expected that our study can provide some guidance for docking-based virtual screening to discover novel type II kinase inhibitors, as well as other allosteric inhibitors.

SUMMARY: Protein-protein interactions (PPIs) have been regarded as an attractive emerging class of therapeutic targets for the development of new treatments. Computational approaches, especially molecular docking, have been extensively employed to predict the binding structures of PPI-inhibitors or discover novel small molecule PPI inhibitors. However, due to the relatively 'undruggable' features of PPI interfaces, accurate predictions of the binding structures for ligands towards PPI targets are quite challenging for most docking algorithms. Here, we constructed a non-redundant pose ranking benchmark dataset for small-molecule PPI inhibitors, which contains 900 binding poses for 184 protein-ligand complexes. Then, we evaluated the performance of MM/PB(GB)SA approaches to identify the correct binding poses for PPI inhibitors, including two Prime MM/GBSA procedures from the Schrödinger suite and seven different MM/PB(GB)SA procedures from the Amber package. Our results showed that MM/PBSA outperformed the Glide SP scoring function (success rate of 58.6%) and MM/GBSA in most cases, especially the PB3 procedure which could achieve an overall success rate of ∼74%. Moreover, the GB6 procedure (success rate of 68.9%) performed much better than the other MM/GBSA procedures, highlighting the excellent potential of the GBNSR6 implicit solvation model for pose ranking. Finally, we developed the webserver of Fast Amber Rescoring for PPI Inhibitors (farPPI), which offers a freely available service to rescore the docking poses for PPI inhibitors by using the MM/PB(GB)SA methods. AVAILABILITY AND IMPLEMENTATION: farPPI web server is freely available at http://cadd.zju.edu.cn/farppi/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.