Constructing and Validating High-Performance MIEC-SVM Models in Virtual Screening for Kinases: A Better Way for Actives Discovery

Inactive-enriched machine-learning models exploiting patent data improve structure-based virtual screening for PDL1 dimerizers

INTRODUCTION

Small-molecule Programmable Cell Death Protein 1/Programmable Death-Ligand 1 (PD1/PDL1) inhibition via PDL1 dimerization has the potential to lead to inexpensive drugs with better cancer patient outcomes and milder side effects. However, this therapeutic approach has proven challenging, with only one PDL1 dimerizer reaching early clinical trials so far. There is hence a need for fast and accurate methods to develop alternative PDL1 dimerizers.

OBJECTIVES

We aim to show that structure-based virtual screening (SBVS) based on PDL1-specific machine-learning (ML) scoring functions (SFs) is a powerful drug design tool for detecting PD1/PDL1 inhibitors via PDL1 dimerization.

METHODS

By incorporating the latest MLSF advances, we generated and evaluated PDL1-specific MLSFs (classifiers and inactive-enriched regressors) on two demanding test sets.

RESULTS

60 PDL1-specific MLSFs (30 classifiers and 30 regressors) were generated. Our large-scale analysis provides highly predictive PDL1-specific MLSFs that benefitted from training with large volumes of docked inactives and enabling inactive-enriched regression.

CONCLUSION

PDL1-specific MLSFs strongly outperformed generic SFs of various types on this target and are released here without restrictions.

Deep learning pipeline for accelerating virtual screening in drug discovery

In the race to combat ever-evolving diseases, the drug discovery process often faces the hurdles of high-cost and time-consuming procedures. To tackle these challenges and enhance the efficiency of identifying new therapeutic agents, we introduce VirtuDockDL, which is a streamlined Python-based web platform utilizing deep learning for drug discovery. This pipeline employs a Graph Neural Network to analyze and predict the effectiveness of various compounds as potential drug candidates. During the validation phase, VirtuDockDL was instrumental in identifying non-covalent inhibitors against the VP35 protein of the Marburg virus, a critical target given the virus's high fatality rate and limited treatment options. Further, in benchmarking, VirtuDockDL achieved 99% accuracy, an F1 score of 0.992, and an AUC of 0.99 on the HER2 dataset, surpassing DeepChem (89% accuracy) and AutoDock Vina (82% accuracy). Compared to RosettaVS, MzDOCK, and PyRMD, VirtuDockDL outperformed them by combining both ligand- and structure-based screening with deep learning. While RosettaVS excels in accurate docking but lacks high-throughput screening, and PyRMD focuses on ligand-based methods without AI integration, VirtuDockDL offers superior predictive accuracy and full automation for large-scale datasets, making it ideal for comprehensive drug discovery workflows. These results underscore the tool's capability to identify high-affinity inhibitors accurately across various targets, including the HER2 protein for cancer therapy, TEM-1 beta-lactamase for bacterial infections, and the CYP51 enzyme for fungal infections like Candidiasis. To sum up, VirtuDockDL combines user-friendly interface design with powerful computational capabilities to facilitate rapid, cost-effective drug discovery and development. The integration of AI in drug discovery could potentially transform the landscape of pharmaceutical research, providing faster responses to global health challenges. The VirtuDockDL is available at https://github.com/FatimaNoor74/VirtuDockDL .

A potent new-scaffold androgen receptor antagonist discovered on the basis of a MIEC-SVM model

Prostate cancer (PCa) is the second most prevalent malignancy among men worldwide. The aberrant activation of androgen receptor (AR) signaling has been recognized as a crucial oncogenic driver for PCa and AR antagonists are widely used in PCa therapy. To develop novel AR antagonist, a machine-learning MIEC-SVM model was established for the virtual screening and 51 candidates were selected and submitted for bioactivity evaluation. To our surprise, a new-scaffold AR antagonist C2 with comparable bioactivity with Enz was identified at the initial round of screening. C2 showed pronounced inhibition on the transcriptional function (IC50 = 0.63 μM) and nuclear translocation of AR and significant antiproliferative and antimetastatic activity on PCa cell line of LNCaP. In addition, C2 exhibited a stronger ability to block the cell cycle of LNCaP than Enz at lower dose and superior AR specificity. Our study highlights the success of MIEC-SVM in discovering AR antagonists, and compound C2 presents a promising new scaffold for the development of AR-targeted therapeutics.

RLpMIEC: High-Affinity Peptide Generation Targeting Major Histocompatibility Complex-I Guided and Interpreted by Interaction Spectrum-Navigated Reinforcement Learning

Major histocompatibility complex (MHC) plays a vital role in presenting epitopes (short peptides from pathogenic proteins) to T-cell receptors (TCRs) to trigger the subsequent immune responses. Vaccine design targeting MHC generally aims to find epitopes with a high binding affinity for MHC presentation. Nevertheless, to find novel epitopes usually requires high-throughput screening of bulk peptide database, which is time-consuming, labor-intensive, more unaffordable, and very expensive. Excitingly, the past several years have witnessed the great success of artificial intelligence (AI) in various fields, such as natural language processing (NLP, e.g., GPT-4), protein structure prediction and engineering (e.g., AlphaFold2), and so on. Therefore, herein, we propose a deep reinforcement-learning (RL)-based generative algorithm, RLpMIEC, to quantitatively design peptide targeting MHC-I systems. Specifically, RLpMIEC combines the energetic spectrum (namely, the molecular interaction energy component, MIEC) based on the peptide-MHC interaction and the sequence information to generate peptides with strong binding affinity and precise MIEC spectra to accelerate the discovery of candidate peptide vaccines. RLpMIEC performs well in all the generative capability evaluations and can generate peptides with strong binding affinities and precise MIECs and, moreover, with high interpretability, demonstrating its powerful capability in participation for accelerating peptide-based vaccine development.

Comprehensive machine learning boosts structure-based virtual screening for PARP1 inhibitors

Poly ADP-ribose polymerase 1 (PARP1) is an attractive therapeutic target for cancer treatment. Machine-learning scoring functions constitute a promising approach to discovering novel PARP1 inhibitors. Cutting-edge PARP1-specific machine-learning scoring functions were investigated using semi-synthetic training data from docking activity-labelled molecules: known PARP1 inhibitors, hard-to-discriminate decoys property-matched to them with generative graph neural networks and confirmed inactives. We further made test sets harder by including only molecules dissimilar to those in the training set. Comprehensive analysis of these datasets using five supervised learning algorithms, and protein-ligand fingerprints extracted from docking poses and ligand only features revealed one highly predictive scoring function. This is the PARP1-specific support vector machine-based regressor, when employing PLEC fingerprints, which achieved a high Normalized Enrichment Factor at the top 1% on the hardest test set (NEF1% = 0.588, median of 10 repetitions), and was more predictive than any other investigated scoring function, especially the classical scoring function employed as baseline.

A practical guide to machine-learning scoring for structure-based virtual screening

Structure-based virtual screening (SBVS) via docking has been used to discover active molecules for a range of therapeutic targets. Chemical and protein data sets that contain integrated bioactivity information have increased both in number and in size. Artificial intelligence and, more concretely, its machine-learning (ML) branch, including deep learning, have effectively exploited these data sets to build scoring functions (SFs) for SBVS against targets with an atomic-resolution 3D model (e.g., generated by X-ray crystallography or predicted by AlphaFold2). Often outperforming their generic and non-ML counterparts, target-specific ML-based SFs represent the state of the art for SBVS. Here, we present a comprehensive and user-friendly protocol to build and rigorously evaluate these new SFs for SBVS. This protocol is organized into four sections: (i) using a public benchmark of a given target to evaluate an existing generic SF; (ii) preparing experimental data for a target from public repositories; (iii) partitioning data into a training set and a test set for subsequent target-specific ML modeling; and (iv) generating and evaluating target-specific ML SFs by using the prepared training-test partitions. All necessary code and input/output data related to three example targets (acetylcholinesterase, HMG-CoA reductase, and peroxisome proliferator-activated receptor-α) are available at https://github.com/vktrannguyen/MLSF-protocol , can be run by using a single computer within 1 week and make use of easily accessible software/programs (e.g., Smina, CNN-Score, RF-Score-VS and DeepCoy) and web resources. Our aim is to provide practical guidance on how to augment training data to enhance SBVS performance, how to identify the most suitable supervised learning algorithm for a data set, and how to build an SF with the highest likelihood of discovering target-active molecules within a given compound library.

Development of Kinase-Centric Drugs: A Computational Perspective

Kinases are prominent drug targets in the pharmaceutical and research community due to their involvement in signal transduction, physiological responses, and upon dysregulation, in diseases such as cancer, neurological and autoimmune disorders. Several FDA-approved small-molecule drugs have been developed to combat human diseases since Gleevec was approved for the treatment of chronic myelogenous leukemia. Kinases were considered "undruggable" in the beginning. Several FDA-approved small-molecule drugs have become available in recent years. Most of these drugs target ATP-binding sites, but a few target allosteric sites. Among kinases that belong to the same family, the catalytic domain shows high structural and sequence conservation. Inhibitors of ATP-binding sites can cause off-target binding. Because members of the same family have similar sequences and structural patterns, often complex relationships between kinases and inhibitors are observed. To design and develop drugs with desired selectivity, it is essential to understand the target selectivity for kinase inhibitors. To create new inhibitors with the desired selectivity, several experimental methods have been designed to profile the kinase selectivity of small molecules. Experimental approaches are often expensive, laborious, time-consuming, and limited by the available kinases. Researchers have used computational methodologies to address these limitations in the design and development of effective therapeutics. Many computational methods have been developed over the last few decades, either to complement experimental findings or to forecast kinase inhibitor activity and selectivity. The purpose of this review is to provide insight into recent advances in theoretical/computational approaches for the design of new kinase inhibitors with the desired selectivity and optimization of existing inhibitors.

Deciphering the Shared and Specific Drug Resistance Mechanisms of Anaplastic Lymphoma Kinase via Binding Free Energy Computation

Anaplastic lymphoma kinase (ALK), a tyrosine receptor kinase, has been proven to be associated with the occurrence of numerous malignancies. Although there have been already at least 3 generations of ALK inhibitors approved by FDA or in clinical trials, the occurrence of various mutations seriously attenuates the effectiveness of the drugs. Unfortunately, most of the drug resistance mechanisms still remain obscure. Therefore, it is necessary to reveal the bottom reasons of the drug resistance mechanisms caused by the mutations. In this work, on the basis of verifying the accuracy of 2 main kinds of binding free energy calculation methodologies [end-point method of Molecular Mechanics with Poisson-Boltzmann/Generalized Born and Surface Area (MM/PB(GB)SA) and alchemical method of Thermodynamic Integration (TI)], we performed a systematic analysis on the ALK systems to explore the underlying shared and specific drug resistance mechanisms, covering the one-drug-multiple-mutation and multiple-drug-one-mutation cases. Through conventional molecular dynamics (cMD) simulation in conjunction with MM/PB(GB)SA and umbrella sampling (US) in conjunction with contact network analysis (CNA), the resistance mechanisms of the in-pocket, out-pocket, and multiple-site mutations were revealed. Especially for the out-pocket mutation, a possible transfer chain of the mutation effect was revealed, and the reason why different drugs exhibited various sensitivities to the same mutation was also uncovered. The proposed mechanisms may be prevalent in various drug resistance cases.

Beware of Simple Methods for Structure-Based Virtual Screening: The Critical Importance of Broader Comparisons

We discuss how data unbiasing and simple methods such as protein-ligand Interaction FingerPrint (IFP) can overestimate virtual screening performance. We also show that IFP is strongly outperformed by target-specific machine-learning scoring functions, which were not considered in a recent report concluding that simple methods were better than machine-learning scoring functions at virtual screening.

Reducing false positive rate of docking-based virtual screening by active learning

Machine learning-based scoring functions (MLSFs) have become a very favorable alternative to classical scoring functions because of their potential superior screening performance. However, the information of negative data used to construct MLSFs was rarely reported in the literature, and meanwhile the putative inactive molecules recorded in existing databases usually have obvious bias from active molecules. Here we proposed an easy-to-use method named AMLSF that combines active learning using negative molecular selection strategies with MLSF, which can iteratively improve the quality of inactive sets and thus reduce the false positive rate of virtual screening. We chose energy auxiliary terms learning as the MLSF and validated our method on eight targets in the diverse subset of DUD-E. For each target, we screened the IterBioScreen database by AMLSF and compared the screening results with those of the four control models. The results illustrate that the number of active molecules in the top 1000 molecules identified by AMLSF was significantly higher than those identified by the control models. In addition, the free energy calculation results for the top 10 molecules screened out by the AMLSF, null model and control models based on DUD-E also proved that more active molecules can be identified, and the false positive rate can be reduced by AMLSF.

Identification of Novel Arachidonic Acid 15-Lipoxygenase Inhibitors Based on the Bayesian Classifier Model and Computer-Aided High-Throughput Virtual Screening

Ferroptosis is an iron-dependent lipid peroxidative form of cell death that is distinct from apoptosis and necrosis. ALOX15, also known as arachidonic acid 15-lipoxygenase, promotes ferroptosis by converting intracellular unsaturated lipids into oxidized lipid intermediates and is an important ferroptosis target. In this study, a naive Bayesian machine learning classifier with a structure-based, high-throughput screening approach and a molecular docking program were combined to screen for three compounds with excellent target-binding potential. In the absorption, distribution, metabolism, excretion, and toxicity characterization, three candidate molecules were predicted to exhibit drug-like properties. The subsequent molecular dynamics simulations confirmed their stable binding to the targets. The findings indicated that the compounds exhibited excellent potential ALOX15 inhibitor capacity, thereby providing novel candidates for the treatment of inflammatory ischemia-related diseases caused by ferroptosis.

Determination of Molecule Category of Ligands Targeting the Ligand-Binding Pocket of Nuclear Receptors with Structural Elucidation and Machine Learning

The mechanism of transcriptional activation/repression of the nuclear receptors (NRs) involves two main conformations of the NR protein, namely, the active (agonistic) and inactive (antagonistic) conformations. Binding of agonists or antagonists to the ligand-binding pocket (LBP) of NRs can regulate the downstream signaling pathways with different physiological effects. However, it is still hard to determine the molecular type of a LBP-bound ligand because both the agonists and antagonists bind to the same position of the protein. Therefore, it is necessary to develop precise and efficient methods to facilitate the discrimination of agonists and antagonists targeting the LBP of NRs. Here, combining structural and energetic analyses with machine-learning (ML) algorithms, we constructed a series of structure-based ML models to determine the molecular category of the LBP-bound ligands. We show that the proposed models work robustly and with high accuracy (ACC > 0.9) for determining the category of molecules derived from docking-based and crystallized poses. Furthermore, the models are also capable of determining the molecular category of ligands with dual opposite functions on different NRs (i.e., working as an agonist in one NR target, whereas functioning as an antagonist in another) with reasonable accuracy. The proposed method is expected to facilitate the determination of the molecular properties of ligands targeting the LBP of NRs with structural interpretation.

fastDRH: a webserver to predict and analyze protein-ligand complexes based on molecular docking and MM/PB(GB)SA computation

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/.

Structure-based protein-ligand interaction fingerprints for binding affinity prediction

Binding affinity prediction (BAP) using protein–ligand complex structures is crucial to computer-aided drug design, but remains a challenging problem. To achieve efficient and accurate BAP, machine-learning scoring functions (SFs) based on a wide range of descriptors have been developed. Among those descriptors, protein–ligand interaction fingerprints (IFPs) are competitive due to their simple representations, elaborate profiles of key interactions and easy collaborations with machine-learning algorithms. In this paper, we have adopted a building-block-based taxonomy to review a broad range of IFP models, and compared representative IFP-based SFs in target-specific and generic scoring tasks. Atom-pair-counts-based and substructure-based IFPs show great potential in these tasks.

Enhancing preclinical drug discovery with artificial intelligence

Artificial intelligence (AI) is becoming an integral part of drug discovery. It has the potential to deliver across the drug discovery and development value chain, starting from target identification and reaching through clinical development. In this review, we provide an overview of current AI technologies and a glimpse of how AI is reimagining preclinical drug discovery by highlighting examples where AI has made a real impact. Considering the excitement and hyperbole surrounding AI in drug discovery, we aim to present a realistic view by discussing both opportunities and challenges in adopting AI in drug discovery.

The impact of cross-docked poses on performance of machine learning classifier for protein-ligand binding pose prediction

Structure-based drug design depends on the detailed knowledge of the three-dimensional (3D) structures of protein-ligand binding complexes, but accurate prediction of ligand-binding poses is still a major challenge for molecular docking due to deficiency of scoring functions (SFs) and ignorance of protein flexibility upon ligand binding. In this study, based on a cross-docking dataset dedicatedly constructed from the PDBbind database, we developed several XGBoost-trained classifiers to discriminate the near-native binding poses from decoys, and systematically assessed their performance with/without the involvement of the cross-docked poses in the training/test sets. The calculation results illustrate that using Extended Connectivity Interaction Features (ECIF), Vina energy terms and docking pose ranks as the features can achieve the best performance, according to the validation through the random splitting or refined-core splitting and the testing on the re-docked or cross-docked poses. Besides, it is found that, despite the significant decrease of the performance for the threefold clustered cross-validation, the inclusion of the Vina energy terms can effectively ensure the lower limit of the performance of the models and thus improve their generalization capability. Furthermore, our calculation results also highlight the importance of the incorporation of the cross-docked poses into the training of the SFs with wide application domain and high robustness for binding pose prediction. The source code and the newly-developed cross-docking datasets can be freely available at https://github.com/sc8668/ml_pose_prediction and https://zenodo.org/record/5525936 , respectively, under an open-source license. We believe that our study may provide valuable guidance for the development and assessment of new machine learning-based SFs (MLSFs) for the predictions of protein-ligand binding poses.

Discovery of potential inhibitors targeting the kinase domain of polynucleotide kinase/phosphatase (PNKP): Homology modeling, virtual screening based on multiple conformations, and molecular dynamics simulation

In recent years, the level of interest has been increased in developing the DNA-repair inhibitors, to enhance the cytotoxic effects in the treatment of cancers. Polynucleotide kinase/phosphatase (PNKP) is a critical human DNA repair enzyme that repairs DNA strand breaks by catalyzing the restoration of 5'-phosphate and 3'-hydroxyl termini that are required for subsequent processing by DNA ligases and polymerases. PNKP is the only protein that repairs the 3'-hydroxyl group and 5'-phosphate group, which depicts PNKP as a potential therapeutic target. Besides, PNKP is the only DNA-repair enzyme that contains the 5'-kinase activity, therefore, targeting this kinase domain would motivate the development of novel PNKP-specific inhibitors. However, there are neither crystal structures of human PNKP nor the kinase inhibitors reported so far. Thus, in this present study, a sequential molecular docking-based virtual screening with multiple PNKP conformations integrating homology modeling, molecular dynamics simulation, and binding free energy calculation was developed to discover novel PNKP kinase inhibitors, and the top-scored molecule was finally submitted to molecular dynamics simulation to reveal the binding mechanism between the inhibitor and PNKP. Taken together, the current study could provide some guidance for the molecular docking based-virtual screening of novel PNKP kinase inhibitors.

Multitask deep networks with grid featurization achieve improved scoring performance for protein-ligand binding

Deep learning-based methods have been extensively developed to improve scoring performance in structure-based drug discovery. Extending multitask deep networks in addressing pharmaceutical problems shows remarkable improvements over single task network. Recently, grid featurization has been introduced to convert protein-ligand complex co-ordinates into fingerprints with the advantage of incorporating inter- and intra-molecular information. The combination of grid featurization with multitask deep networks would hold great potential to boost the scoring performance. We examined the performance of three novel multitask deep networks (standard multitask, bypass, and progressive network) in reproducing the binding affinities of protein-ligand complexes in comparison with AutoDock Vina docking and MM/GBSA method. Among five evaluated methods, progressive network combined with grid featurization provided the best Pearson correlation coefficient (0.74) and least mean absolute average error (0.98) for the overall scoring performance. Moreover, all networks increased screening ability for the re-docking pose and progressive network even achieved AUC of 0.87 over 0.52 of AutoDock Vina. Our results demonstrated that progressive network combined with grid featurization would be one powerful rescoring approach to strengthen screening results after obtaining protein-ligand complex in the conventional docking software.

MolAICal: a soft tool for 3D drug design of protein targets by artificial intelligence and classical algorithm

Deep learning is an important branch of artificial intelligence that has been successfully applied into medicine and two-dimensional ligand design. The three-dimensional (3D) ligand generation in the 3D pocket of protein target is an interesting and challenging issue for drug design by deep learning. Here, the MolAICal software is introduced to supply a way for generating 3D drugs in the 3D pocket of protein targets by combining with merits of deep learning model and classical algorithm. The MolAICal software mainly contains two modules for 3D drug design. In the first module of MolAICal, it employs the genetic algorithm, deep learning model trained by FDA-approved drug fragments and Vinardo score fitting on the basis of PDBbind database for drug design. In the second module, it uses deep learning generative model trained by drug-like molecules of ZINC database and molecular docking invoked by Autodock Vina automatically. Besides, the Lipinski's rule of five, Pan-assay interference compounds (PAINS), synthetic accessibility (SA) and other user-defined rules are introduced for filtering out unwanted ligands in MolAICal. To show the drug design modules of MolAICal, the membrane protein glucagon receptor and non-membrane protein SARS-CoV-2 main protease are chosen as the investigative drug targets. The results show MolAICal can generate the various and novel ligands with good binding scores and appropriate XLOGP values. We believe that MolAICal can use the advantages of deep learning model and classical programming for designing 3D drugs in protein pocket. MolAICal is freely for any nonprofit purpose and accessible at https://molaical.github.io.

Recent progress on the prospective application of machine learning to structure-based virtual screening

As more bioactivity and protein structure data become available, scoring functions (SFs) using machine learning (ML) to leverage these data sets continue to gain further accuracy and broader applicability. Advances in our understanding of the optimal ways to train and evaluate these ML-based SFs have introduced further improvements. One of these advances is how to select the most suitable decoys (molecules assumed inactive) to train or test an ML-based SF on a given target. We also review the latest applications of ML-based SFs for prospective structure-based virtual screening (SBVS), with a focus on the observed improvement over those using classical SFs. Finally, we provide recommendations for future prospective SBVS studies based on the findings of recent methodological studies.

Artificial intelligence and machine learning-aided drug discovery in central nervous system diseases: State-of-the-arts and future directions

Neurological disorders significantly outnumber diseases in other therapeutic areas. However, developing drugs for central nervous system (CNS) disorders remains the most challenging area in drug discovery, accompanied with the long timelines and high attrition rates. With the rapid growth of biomedical data enabled by advanced experimental technologies, artificial intelligence (AI) and machine learning (ML) have emerged as an indispensable tool to draw meaningful insights and improve decision making in drug discovery. Thanks to the advancements in AI and ML algorithms, now the AI/ML-driven solutions have an unprecedented potential to accelerate the process of CNS drug discovery with better success rate. In this review, we comprehensively summarize AI/ML-powered pharmaceutical discovery efforts and their implementations in the CNS area. After introducing the AI/ML models as well as the conceptualization and data preparation, we outline the applications of AI/ML technologies to several key procedures in drug discovery, including target identification, compound screening, hit/lead generation and optimization, drug response and synergy prediction, de novo drug design, and drug repurposing. We review the current state-of-the-art of AI/ML-guided CNS drug discovery, focusing on blood-brain barrier permeability prediction and implementation into therapeutic discovery for neurological diseases. Finally, we discuss the major challenges and limitations of current approaches and possible future directions that may provide resolutions to these difficulties.

Machine learning on ligand-residue interaction profiles to significantly improve binding affinity prediction

Structure-based virtual screenings (SBVSs) play an important role in drug discovery projects. However, it is still a challenge to accurately predict the binding affinity of an arbitrary molecule binds to a drug target and prioritize top ligands from an SBVS. In this study, we developed a novel method, using ligand-residue interaction profiles (IPs) to construct machine learning (ML)-based prediction models, to significantly improve the screening performance in SBVSs. Such a kind of the prediction model is called an IP scoring function (IP-SF). We systematically investigated how to improve the performance of IP-SFs from many perspectives, including the sampling methods before interaction energy calculation and different ML algorithms. Using six drug targets with each having hundreds of known ligands, we conducted a critical evaluation on the developed IP-SFs. The IP-SFs employing a gradient boosting decision tree (GBDT) algorithm in conjunction with the MIN + GB simulation protocol achieved the best overall performance. Its scoring power, ranking power and screening power significantly outperformed the Glide SF. First, compared with Glide, the average values of mean absolute error and root mean square error of GBDT/MIN + GB decreased about 38 and 36%, respectively. Second, the mean values of squared correlation coefficient and predictive index increased about 225 and 73%, respectively. Third, more encouragingly, the average value of the areas under the curve of receiver operating characteristic for six targets by GBDT, 0.87, is significantly better than that by Glide, which is only 0.71. Thus, we expected IP-SFs to have broad and promising applications in SBVSs.

Property-Unmatched Decoys in Docking Benchmarks

Enrichment of ligands versus property-matched decoys is widely used to test and optimize docking library screens. However, the unconstrained optimization of enrichment alone can mislead, leading to false confidence in prospective performance. This can arise by over-optimizing for enrichment against property-matched decoys, without considering the full spectrum of molecules to be found in a true large library screen. Adding decoys representing charge extrema helps mitigate over-optimizing for electrostatic interactions. Adding decoys that represent the overall characteristics of the library to be docked allows one to sample molecules not represented by ligands and property-matched decoys but that one will encounter in a prospective screen. An optimized version of the DUD-E set (DUDE-Z), as well as Extrema and sets representing broad features of the library (Goldilocks), is developed here. We also explore the variability that one can encounter in enrichment calculations and how that can temper one's confidence in small enrichment differences. The new tools and new decoy sets are freely available at http://tldr.docking.org and http://dudez.docking.org.

Generating Property-Matched Decoy Molecules Using Deep Learning

MOTIVATION

An essential step in the development of virtual screening methods is the use of established sets of actives and decoys for benchmarking and training. However, the decoy molecules in commonly used sets are biased meaning that methods often exploit these biases to separate actives and decoys, and do not necessarily learn to perform molecular recognition. This fundamental issue prevents generalisation and hinders virtual screening method development.

RESULTS

We have developed a deep learning method (DeepCoy) that generates decoys to a user's preferred specification in order to remove such biases or construct sets with a defined bias.We validated DeepCoy using two established benchmarks, DUD-E and DEKOIS 2.0. For all 102 DUD-E targets and 80 of the 81 DEKOIS 2.0 targets, our generated decoy molecules more closely matched the active molecules' physicochemical properties while introducing no discernible additional risk of false negatives. The DeepCoy decoys improved the Deviation from Optimal Embedding (DOE) score by an average of 81% and 66%, respectively, decreasing from 0.166 to 0.032 for DUD-E and from 0.109 to 0.038 for DEKOIS 2.0. Further, the generated decoys are harder to distinguish than the original decoy molecules via docking with Autodock Vina, with virtual screening performance falling from an AUC ROC of 0.70 to 0.63.

AVAILABILITY AND IMPLEMENTATION

The code is available at https://github.com/oxpig/DeepCoy. Generated molecules can be downloaded from http://opig.stats.ox.ac.uk/resources.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Comparative Study of Interactions between Human cGAS and Inhibitors: Insights from Molecular Dynamics and MM/PBSA Studies

Recent studies have identified cyclic GMP-AMP synthase (cGAS) as an important target for treating autoimmune diseases, and several inhibitors of human cGAS (hcGAS) and their structures in complexation with hcGAS have been reported. However, the mechanisms via which these inhibitors interact with hcGAS are not completely understood. Here, we aimed to assess the performance of molecular mechanics/Poisson–Boltzmann solvent-accessible surface area (MM/PBSA) in evaluating the binding affinity of various hcGAS inhibitors and to elucidate their detailed interactions with hcGAS from an energetic viewpoint. Using molecular dynamics (MD) simulation and MM/PBSA approaches, the estimated free energies were in good agreement with the experimental ones, with a Pearson’s correlation coefficient and Spearman’s rank coefficient of 0.67 and 0.46, respectively. In per-residue energy decomposition analysis, four residues, K362, R376, Y436, and K439 in hcGAS were found to contribute significantly to the binding with inhibitors via hydrogen bonding, salt bridges, and various π interactions, such as π· · ·π stacking, cation· · ·π, hydroxyl· · ·π, and alkyl· · ·π interactions. In addition, we discussed other key interactions between specific residues and ligands, in particular, between H363 and JUJ, F379 and 9BY, and H437 and 8ZM. The sandwiched structures of the inhibitor bound to the guanidinium group of R376 and the phenyl ring of Y436 were also consistent with the experimental data. The results indicated that MM/PBSA in combination with other virtual screening methods, could be a reliable approach to discover new hcGAS inhibitors and thus is valuable for potential treatments of cGAS-dependent inflammatory diseases.

Accuracy or novelty: what can we gain from target-specific machine-learning-based scoring functions in virtual screening?

Machine-learning (ML)-based scoring functions (MLSFs) have gradually emerged as a promising alternative for protein-ligand binding affinity prediction and structure-based virtual screening. However, clouds of doubts have still been raised against the benefits of this novel type of scoring functions (SFs). In this study, to benchmark the performance of target-specific MLSFs on a relatively unbiased dataset, the MLSFs trained from three representative protein-ligand interaction representations were assessed on the LIT-PCBA dataset, and the classical Glide SP SF and three types of ligand-based quantitative structure-activity relationship (QSAR) models were also utilized for comparison. Two major aspects in virtual screening campaigns, including prediction accuracy and hit novelty, were systematically explored. The calculation results illustrate that the tested target-specific MLSFs yielded generally superior performance over the classical Glide SP SF, but they could hardly outperform the 2D fingerprint-based QSAR models. Although substantial improvements could be achieved by integrating multiple types of protein-ligand interaction features, the MLSFs were still not sufficient to exceed MACCS-based QSAR models. In terms of the correlations between the hit ranks or the structures of the top-ranked hits, the MLSFs developed by different featurization strategies would have the ability to identify quite different hits. Nevertheless, it seems that target-specific MLSFs do not have the intrinsic attributes of a traditional SF and may not be a substitute for classical SFs. In contrast, MLSFs can be regarded as a new derivative of ligand-based QSAR models. It is expected that our study may provide valuable guidance for the assessment and further development of target-specific MLSFs.

Merging Ligand-Based and Structure-Based Methods in Drug Discovery: An Overview of Combined Virtual Screening Approaches

Virtual screening (VS) is an outstanding cornerstone in the drug discovery pipeline. A variety of computational approaches, which are generally classified as ligand-based (LB) and structure-based (SB) techniques, exploit key structural and physicochemical properties of ligands and targets to enable the screening of virtual libraries in the search of active compounds. Though LB and SB methods have found widespread application in the discovery of novel drug-like candidates, their complementary natures have stimulated continued efforts toward the development of hybrid strategies that combine LB and SB techniques, integrating them in a holistic computational framework that exploits the available information of both ligand and target to enhance the success of drug discovery projects. In this review, we analyze the main strategies and concepts that have emerged in the last years for defining hybrid LB + SB computational schemes in VS studies. Particularly, attention is focused on the combination of molecular similarity and docking, illustrating them with selected applications taken from the literature.

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.

The impact of compound library size on the performance of scoring functions for structure-based virtual screening

Larger training datasets have been shown to improve the accuracy of machine learning (ML)-based scoring functions (SFs) for structure-based virtual screening (SBVS). In addition, massive test sets for SBVS, known as ultra-large compound libraries, have been demonstrated to enable the fast discovery of selective drug leads with low-nanomolar potency. This proof-of-concept was carried out on two targets using a single docking tool along with its SF. It is thus unclear whether this high level of performance would generalise to other targets, docking tools and SFs. We found that screening a larger compound library results in more potent actives being identified in all six additional targets using a different docking tool along with its classical SF. Furthermore, we established that a way to improve the potency of the retrieved molecules further is to rank them with more accurate ML-based SFs (we found this to be true in four of the six targets; the difference was not significant in the remaining two targets). A 3-fold increase in average hit rate across targets was also achieved by the ML-based SFs. Lastly, we observed that classical and ML-based SFs often find different actives, which supports using both types of SFs on those targets.

Improving structure-based virtual screening performance via learning from scoring function components

Scoring functions (SFs) based on complex machine learning (ML) algorithms have gradually emerged as a promising alternative to overcome the weaknesses of classical SFs. However, extensive efforts have been devoted to the development of SFs based on new protein-ligand interaction representations and advanced alternative ML algorithms instead of the energy components obtained by the decomposition of existing SFs. Here, we propose a new method named energy auxiliary terms learning (EATL), in which the scoring components are extracted and used as the input for the development of three levels of ML SFs including EATL SFs, docking-EATL SFs and comprehensive SFs with ascending VS performance. The EATL approach not only outperforms classical SFs for the absolute performance (ROC) and initial enrichment (BEDROC) but also yields comparable performance compared with other advanced ML-based methods on the diverse subset of Directory of Useful Decoys: Enhanced (DUD-E). The test on the relatively unbiased actives as decoys (AD) dataset also proved the effectiveness of EATL. Furthermore, the idea of learning from SF components to yield improved screening power can also be extended to other docking programs and SFs available.

Beware of the generic machine learning-based scoring functions in structure-based virtual screening

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.

Glucose-6-Phosphate Upregulates Txnip Expression by Interacting With MondoA

The major metabolic fates of glucose in cells are glycolysis and the pentose phosphate pathway, and they share the first step: converting glucose to glucose-6-phosphate (G6P). Here, we show that G6P can be sensed by the transcription factor MondoA/Mlx to modulate Txnip expression. Endogenous knockdown and EMSA (gel migration assay) analyses both confirmed that G6P is the metabolic intermediate that activates the heterocomplex MondoA/Mlx to elicit the expression of Txnip. Additionally, the three-dimensional structure of MondoA is modeled, and the binding mode of G6P to MondoA is also predicted by in silico molecular docking and binding free energy calculation. Finally, free energy decomposition and mutational analyses suggest that certain residues in MondoA, GKL139-141 in particular, mediate its binding with G6P to activate MondoA, which signals the upregulation of the expression of Txnip.

Combined strategies in structure-based virtual screening

The identification and optimization of lead compounds are inalienable components in drug design and discovery pipelines. As a powerful computational approach for the identification of hits with novel structural scaffolds, structure-based virtual screening (SBVS) has exhibited a remarkably increasing influence in the early stages of drug discovery. During the past decade, a variety of techniques and algorithms have been proposed and tested with different purposes in the scope of SBVS. Although SBVS has been a common and proven technology, it still shows some challenges and problems that are needed to be addressed, where the negative influence regardless of protein flexibility and the inaccurate prediction of binding affinity are the two major challenges. Here, focusing on these difficulties, we summarize a series of combined strategies or workflows developed by our group and others. Furthermore, several representative successful applications from recent publications are also discussed to demonstrate the effectiveness of the combined SBVS strategies in drug discovery campaigns.

Identification of novel peptidomimetics targeting the polo-box domain of polo-like kinase 1

A series of new peptidomimetics targeting the polo-box domain (PBD) of polo-like kinase 1 (Plk1) was identified based on the potent and selective pentapeptide Plk1 PBD inhibitor PLHSpT. Unnatural amino acid residues were introduced to the newly designed compound and the N-terminal substituent of the peptidomimetic was investigated. The optimized compound 9 inhibited the Plk1 PBD with IC50 of 0.267 μM and showed almost no inhibition to Plk2 PBD or Plk3 PBD at 100 μM. Biolayer interferometry studies demonstrated that compound 9 showed potent binding affinity to Plk1 with a Kd value of 0.164 μM, while no Kd were detected against Plk2 and Plk3. Compound 9 showed improved stability in rat plasma compared to PLHSpT. Binding mode analysis was performed and in agreement with the observed experimental results. There are only two natural amino acids remained in the chemical structure of 9. This study may provide new information for further research on Plk1 PBD inhibitors.

Discovery of novel natural compound inhibitors targeting estrogen receptor α by an integrated virtual screening strategy

Estrogen receptor (ER) is a nuclear hormone receptor and plays an important role in mediating the cellular effects of estrogen. ER can be classified into two receptors: estrogen receptor alpha (ERα) and beta (ERβ), and the former is expressed in 50~80% of breast tumors and has been extensively investigated in breast cancer for decades. Excessive exposure to estrogen can obviously stimulate the growth of breast cancers primarily mediated by ERα, and thus anti-estrogen therapies by small molecules are of concern to clinicians and pharmaceutical industry in the treatment of ERα-positive breast cancers. Although a series of estrogen receptor modulators have been developed, these drugs can lead to resistance and side effects. Therefore, the development of small molecule inhibitors with high target specificity has been intensified. In this pursuit, an integrated computer-aided virtual screening technique, including molecular docking and pharmacophore model screening, was used to screen traditional Chinese medicine (TCM) databases. The compounds with high docking scores and fit values were subjected to ADME (adsorption, distribution, metabolism, excretion) and toxicity prediction, and ten hits were identified as potential inhibitors targeting ERα. Molecular docking was used to investigate the binding modes between ERα and three most potent hits, and molecular dynamic simulations were chosen to explore the stability of these complexes. The rank of the predicted binding free energies evaluated by MM/GBSA is consistent with the docking score. These novel scaffolds discovered in the present study can be used as critical starting point in the drug discovery process for treating ERα-positive breast cancer. Graphical abstract .

Concepts of Artificial Intelligence for Computer-Assisted Drug Discovery

Artificial intelligence (AI), and, in particular, deep learning as a subcategory of AI, provides opportunities for the discovery and development of innovative drugs. Various machine learning approaches have recently (re)emerged, some of which may be considered instances of domain-specific AI which have been successfully employed for drug discovery and design. This review provides a comprehensive portrayal of these machine learning techniques and of their applications in medicinal chemistry. After introducing the basic principles, alongside some application notes, of the various machine learning algorithms, the current state-of-the art of AI-assisted pharmaceutical discovery is discussed, including applications in structure- and ligand-based virtual screening, de novo drug design, physicochemical and pharmacokinetic property prediction, drug repurposing, and related aspects. Finally, several challenges and limitations of the current methods are summarized, with a view to potential future directions for AI-assisted drug discovery and design.

The compromise of virtual screening and its impact on drug discovery

Introduction: Docking and structure-based virtual screening (VS) have been standard approaches in structure-based design for over two decades. However, our understanding of the limitations, potential, and strength of these techniques has enhanced, raising expectations. Areas covered: Based on a survey of reports in the past five years, we assess whether VS: (1) predicts binding poses in agreement with crystallographic data (when available); (2) is a superior screening tool, as often claimed; (3) is successful in identifying chemical scaffolds that can be starting points for subsequent lead optimization cycles. Data shows that knowledge of the target and its chemotypes in postprocessing lead to viable hits in early drug discovery endeavors. Expert opinion: VS is capable of accurate placements in the pocket for the most part, but does not consistently score screening collections accurately. What matters is capitalization on available resources to get closer to a viable lead or optimizable series. Integration of approaches, subjective hit selection guided by knowledge of the receptor or endogenous ligand, libraries driven by experimental guides, validation studies to identify the best docking/scoring that reproduces experimental findings, constraints regarding receptor-ligand interactions, thoroughly designed methodologies, and predefined cutoff scoring criteria strengthen VS's position in pharmaceutical research.

Classical scoring functions for docking are unable to exploit large volumes of structural and interaction data

Motivation

Studies have shown that the accuracy of random forest (RF)-based scoring functions (SFs), such as RF-Score-v3, increases with more training samples, whereas that of classical SFs, such as X-Score, does not. Nevertheless, the impact of the similarity between training and test samples on this matter has not been studied in a systematic manner. It is therefore unclear how these SFs would perform when only trained on protein-ligand complexes that are highly dissimilar or highly similar to the test set. It is also unclear whether SFs based on machine learning algorithms other than RF can also improve accuracy with increasing training set size and to what extent they learn from dissimilar or similar training complexes.

Results

We present a systematic study to investigate how the accuracy of classical and machine-learning SFs varies with protein-ligand complex similarities between training and test sets. We considered three types of similarity metrics, based on the comparison of either protein structures, protein sequences or ligand structures. Regardless of the similarity metric, we found that incorporating a larger proportion of similar complexes to the training set did not make classical SFs more accurate. In contrast, RF-Score-v3 was able to outperform X-Score even when trained on just 32% of the most dissimilar complexes, showing that its superior performance owes considerably to learning from dissimilar training complexes to those in the test set. In addition, we generated the first SF employing Extreme Gradient Boosting (XGBoost), XGB-Score, and observed that it also improves with training set size while outperforming the rest of SFs. Given the continuous growth of training datasets, the development of machine-learning SFs has become very appealing.

Availability and implementation

https://github.com/HongjianLi/MLSF

Supplementary information

Supplementary data are available at Bioinformatics online.

Covalent vs. Non-Covalent Inhibition: Tackling Drug Resistance in EGFR - A Thorough Dynamic Perspective

A persistent challenge in the treatment of non-small cell lung cancer (NSCLC) with EGFR is the emergence of drug-resistant caused by somatic mutations. The EGFR L858R/T790 M double mutant (EGFR^DM ) was found to be the most alarming variant. Despite the development of a wide range of inhibitors, none of them could inhibit EGFR^DM effectively. Recently, 11h and 45a, have been found to be potent inhibitors against EGFR^DM through two distinctive mechanisms, non-covalent and covalent binding, respectively. However, the structural and dynamic implications of the two modes of inhibitions remain unexplored. Herein, two molecular dynamics simulation protocols, coupled with free-energy calculations, were applied to gain insight into the atomistic nature of each binding mode. The comparative analysis confirmed that there is a significant difference in the binding free energy between 11h and 45a (ΔΔG_bind =-21.17 kcal/mol). The main binding force that governs the binding of both inhibitors is vdW, with a higher contribution for 45a. Two residues ARG841 and THR854 were found to have curtailed role in the binding of 45a to EGFR^DM by stabilizing its flexible alcohol chain. The 45a binding to EGFR^DM induces structural rearrangement in the active site to allow easier accessibility of 45a to target residue CYS797. The findings of this work can substantially shed light on new strategies for developing novel classes of covalent and non-covalent inhibitors with increased specificity and potency.

Building Machine-Learning Scoring Functions for Structure-Based Prediction of Intermolecular Binding Affinity

Molecular docking enables large-scale prediction of whether and how small molecules bind to a macromolecular target. Machine-learning scoring functions are particularly well suited to predict the strength of this interaction. Here we describe how to build RF-Score, a scoring function utilizing the machine-learning technique known as Random Forest (RF). We also point out how to use different data, features, and regression models using either R or Python programming languages.

Identification of novel and selective non-peptide inhibitors targeting the polo-box domain of polo-like kinase 1

A series of non-peptide inhibitors targeting the polo-box domain (PBD) of polo-like kinase 1 (Plk1) was designed based on the potent and selective minimal tripeptide Plk1 PBD inhibitor. Seven compounds were designed, synthesized and evaluated for fluorescence polarization (FP) assay. The most promising compound 10 bound to Plk1 PBD with IC50 of 3.37 μM and had no binding to Plk2 PBD or Plk3 PBD at 100 μM. Molecular docking study was performed and possible binding mode was proposed. MM/GBSA binding free energy calculation were in agreement with the observed experimental results. These novel non-peptide selective Plk1 PBD inhibitors provided new lead compounds for further optimization.

Exploring the binding mechanisms of PDE5 with chromeno[2,3-c]pyrrol-9(2H)-one by theoretical approaches

Cyclic nucleotide phosphodiesterase type 5 (PDE5), exclusively specific for the cyclic guanosine monophosphate (cGMP), is an important drug target for the treatment of erectile dysfunction and pulmonary arterial hypertension (PAH). Although many PDE5 inhibitors have been approved, such as sildenafil, vardenafil, tadalafil and so on, extensive studies have reported some side effects, such as vision disturbance and hearing loss as a result of the amino acid sequence and the secondary structural similarity of other PDEs to the catalytic domain of PDE5. In this study, multiple docking strategies, molecular dynamics (MD) simulations, free energy calculations and decomposition were employed to explore the structural determinants of PDE5 with a series of chromeno[2,3-c]pyrrol-9(2H)-one derivatives. First, reliable docking results were obtained using quantum mechanics/molecular mechanics (QM/MM) docking. Then, MD simulations and MM/GBSA free energy calculations were used to explore the dynamic binding process and characterize the binding modes of the inhibitors with different activities. The predicted binding free energies are in good agreement with the experimental data, and the MM/GBSA free energy decomposition analysis sheds light on the importance of hydrogen bonds with Gln817, π-π stacks against Phe820 and hydrophobic residues for the PDE5 binding of the studied inhibitors. The structural and energetic insights obtained here are useful for understanding the molecular mechanism of ligand binding and designing novel potent and selective PDE5 inhibitors with new scaffolds.

Assessing the performance of MM/PBSA and MM/GBSA methods. 7. Entropy effects on the performance of end-point binding free energy calculation approaches

Entropy effects play an important role in drug-target interactions, but the entropic contribution to ligand-binding affinity is often neglected by end-point binding free energy calculation methods, such as MM/GBSA and MM/PBSA, due to the expensive computational cost of normal mode analysis (NMA). Here, we systematically investigated entropy effects on the prediction power of MM/GBSA and MM/PBSA using >1500 protein-ligand systems and six representative AMBER force fields. Two computationally efficient methods, including NMA based on truncated structures and the interaction entropy approach, were used to estimate the entropic contributions to ligand-target binding free energies. In terms of the overall accuracy, we found that, for the minimized structures, in most cases the inclusion of the conformational entropies predicted by truncated NMA (enthalpynmode_min_9Å) compromises the overall accuracy of MM/GBSA and MM/PBSA compared with the enthalpies calculated based on the minimized structures (enthalpymin). However, for the MD trajectories, the binding free energies can be improved by the inclusion of the conformation entropies predicted by either truncated-NMA for a relatively high dielectric constant (εin = 4) or the interaction entropy method for εin = 1-4. In terms of reproducing the absolute binding free energies, the binding free energies estimated by including the truncated-NMA entropies based on the MD trajectories (ΔGnmode_md_9Å) give the lowest average absolute deviations against the experimental data among all the tested strategies for both MM/GBSA and MM/PBSA. Although the inclusion of the truncated NMA based on the MD trajectories (ΔGnmode_md_9Å) for a relatively high dielectric constant gave the overall best result and the lowest average absolute deviations against the experimental data (for the ff03 force field), it needs too much computational time. Alternatively, considering that the interaction entropy method does not incur any additional computational cost and can give comparable (at high dielectric constant, εin = 4) or even better (at low dielectric constant, εin = 1-2) results than the truncated-NMA entropy (ΔGnmode_md_9Å), the interaction entropy approach is recommended to estimate the entropic component for MM/GBSA and MM/PBSA based on MD trajectories, especially for a diverse dataset. Furthermore, we compared the predictions of MM/GBSA with six different AMBER force fields. The results show that the ff03 force field (ff03 for proteins and gaff with AM1-BCC charges for ligands) performs the best, but the predictions given by the tested force fields are comparable, implying that the MM/GBSA predictions are not very sensitive to force fields.