Modeling the protonation states of β-secretase binding pocket by molecular dynamics simulations and docking studies

β-secretase (BACE1) is an aspartyl protease that processes the β-amyloid peptide in the human brain in patients with Alzheimer’s disease. There are two catalytic aspartates (ASP32 and ASP228) in the active domain of BACE1. Although it is believed that the net charge of the Asp dyad is −1, the exact protonation state still remains a matter of debate. We carried out molecular dynamic (MD) simulations for the four protonation states of BACE1 proteins. We applied Glide docking studies to 21 BACE1 inhibitors against the MD extracted conformations. The dynamic results infer that the protein/ligand complex remains stable during the entire simulation course for HD32D228 model. The results show that the hydrogen bonds between the inhibitor and the Asp dyad are maintained in the 10,000th ps snapshot of HD32D228 model. Our results also reveal the significant loop residues in maintaining the active binding conformation in the HD32D228 model. Molecular docking results show that the HD32D228 model provided the best enrichment factor score, suggesting that this model was able to recognize the most active compounds. Our observations provide an evidence for the preference of the anionic state (HD32D228) in BACE1 binding site and are in accord with reported computational data. The protonation state study would provide significant information to assign the correct protonation state for structure-based drug design and docking studies targeting the BACE1 proteins as a tactic to develop potential AD inhibitors.     

Docking studies of matrix metalloproteinase inhibitors: zinc parameter optimization to improve the binding free energy prediction

Docking of metalloproteinase inhibitors remains a challenge due to the zinc multiple coordination geometries and the lack of appropriate force field parameters to model the metal/ligand interactions. In this study, we explore the docking accuracy and scoring reliability for the docking of matrix metalloproteinase (MMP) inhibitors using AutoDock 3.0. Potential problems associated with zinc ion were investigated by docking 16 matrix metalloproteinase ligands to their crystal structures. A good coordination between the zinc binding group (ZBG) and the zinc was shown to be a prerequisite for the ligand to fit the binding site. A simplex optimization of zinc parameters, including zinc radius, well depth, and zinc charges, was performed utilizing the 14 MMP complexes with good docking. The use of optimized zinc parameters (zinc radius: 0.87 A; well depth: 0.35 kcal/mol; and zinc charges: +0.95 e) shows improvement in both docking accuracy at the zinc binding site and the prediction of binding free energies. Although further improvement in the docking procedure, particularly the scoring function is needed, optimization of zinc parameters provides an efficient way to improve the performance of AutoDock as a drug discovery tool.     

New generation of docking programs: Supercomputer validation of force fields and quantum-chemical methods for docking

Discovery of new inhibitors of the protein associated with a given disease is the initial and most important stage of the whole process of the rational development of new pharmaceutical substances. New inhibitors block the active site of the target protein and the disease is cured. Computer-aided molecular modeling can considerably increase effectiveness of new inhibitors development. Reliable predictions of the target protein inhibition by a small molecule, ligand, is defined by the accuracy of docking programs. Such programs position a ligand in the target protein and estimate the protein-ligand binding energy. Positioning accuracy of modern docking programs is satisfactory. However, the accuracy of binding energy calculations is too low to predict good inhibitors. For effective application of docking programs to new inhibitors development the accuracy of binding energy calculations should be higher than 1 kcal/mol. Reasons of limited accuracy of modern docking programs are discussed. One of the most important aspects limiting this accuracy is imperfection of protein-ligand energy calculations. Results of supercomputer validation of several force fields and quantum-chemical methods for docking are presented. The validation was performed by quasi-docking as follows. First, the low energy minima spectra of 16 protein-ligand complexes were found by exhaustive minima search in the MMFF94 force field. Second, energies of the lowest 8192 minima are recalculated with CHARMM force field and PM6-D3H4X and PM7 quantum-chemical methods for each complex. The analysis of minima energies reveals the docking positioning accuracies of the PM7 and PM6-D3H4X quantum-chemical methods and the CHARMM force field are close to one another and they are better than the positioning accuracy of the MMFF94 force field.     

A practical approach to docking of zinc metalloproteinase inhibitors

Forty zinc-dependent metalloproteinase/ligand complexes with known crystal structures were re-docked using five docking/scoring approaches (DOCK, FlexX, DrugScore, GOLD, and AutoDock). Correct geometry of the coordination bonds between the ligand's zinc binding group (ZBG) and the catalytic zinc is important for docking accuracy and scoring reliability. More than 75% of docked poses with RMSD less than 2A were found to have appropriate ZBG binding, but for poor ZBG binding, about 95% of poses failed to dock correctly. Elimination of poses with inappropriate zinc binding resulted in better binding energy predictions that were further improved by dividing the ligands into subsets according to the ZBG (carboxylates, hydroxamates, and phosphorus containing groups). After a subset re-scoring using the regression functions obtained for individual subsets, DrugScore was able to explain 77% and the consensus scoring scheme X-CSCORE even 88% of variance in binding energies. The approach combining ZBG-based pose selection and subset re-scoring improved the hit rate in virtual screening for metalloproteinase inhibitors for all tested methods by 4-16%.     

Protein tyrosine phosphatases: Ligand interaction analysis and optimisation of virtual screening

Docking-based virtual screening is an established component of structure-based drug discovery. Nevertheless, scoring and ranking of computationally docked ligand libraries still suffer from many false positives. Identifying optimal docking parameters for a target protein prior to virtual screening can improve experimental hit rates. Here, we examine protocols for virtual screening against the important but challenging class of drug target, protein tyrosine phosphatases. In this study, common interaction features were identified from analysis of protein–ligand binding geometries of more than 50 complexed phosphatase crystal structures. It was found that two interactions were consistently formed across all phosphatase inhibitors: (1) a polar contact with the conserved arginine residue, and (2) at least one interaction with the P-loop backbone amide. In order to investigate the significance of these features on phosphatase-ligand binding, a series of seeded virtual screening experiments were conducted on three phosphatase enzymes, PTP1B, Cdc25b and IF2. It was observed that when the conserved arginine and P-loop amide interactions were used as pharmacophoric constraints during docking, enrichment of the virtual screen significantly increased in the three studied phosphatases, by up to a factor of two in some cases. Additionally, the use of such pharmacophoric constraints considerably improved the ability of docking to predict the inhibitor's bound pose, decreasing RMSD to the crystallographic geometry by 43% on average. Constrained docking improved enrichment of screens against both open and closed conformations of PTP1B. Incorporation of an ordered water molecule in PTP1B screening was also found to generally improve enrichment. The knowledge-based computational strategies explored here can potentially inform structure-based design of new phosphatase inhibitors using docking-based virtual screening.     

Structure-based design of hERG-neutral antihypertensive oxazalone and imidazolone derivatives

Angiotensin II receptor type 1 (AT1) antagonists are the most recent drug class against hypertension. Recently first crystal structure of AT1 receptor is deposited to the protein data bank (PDB ID: 4YAY). In this work, several molecular screening methods such as molecular docking and de novo design studies were performed and it is found that oxazolone and imidazolone derivatives reveal similar/better interaction energy profiles compared to the FDA approved sartan molecules at the binding site of the AT1 receptor. A database consisting of 3500-fragments were used to enumerate de novo designed imidazolone and oxazolone derivatives and hereby more than 50000 novel small molecules were generated. These derivatives were then used in high throughput virtual screening simulations (Glide/HTVS) to find potent hit molecules. In addition, virtual screening of around 18 million small drug-like compounds from ZINC database were screened at the binding pocket of the AT1 receptor via Glide/HTVS method. Filtered structures were then used in more sophisticated molecular docking simulations protocols (i.e., Glide/SP; Glide/XP; Glide/IFD; Glide/QPLD, and GOLD). However, the K⁺ ion channel/drug interactions should also be considered in studies implemented in molecular level against their cardiovascular risks. Thus, selected compounds with high docking scores via all diverse docking algorithms are also screened at the pore domain regions of human ether-a-go-go-related gene (hERG1) K⁺ channel to remove the high affinity hERG1 blocking compounds. High docking scored compounds at the AT1 with low hERG1 affinity is considered for long molecular dynamics (MD) simulations. Post-processing analysis of MD simulations assisted for better understanding of molecular mechanism of studied compounds at the binding cavity of AT1 receptor. Results of this study can be useful for designing of novel and safe AT1 inhibitors.     

Differential binding mode of diverse cyclooxygenase inhibitors

Non-steroidal anti-inflammatory drugs (NSAIDs) are competitive inhibitors of cyclooxygenase (COX), the enzyme that mediates biosynthesis of prostaglandins and thromboxanes from arachidonic acid. There are at least two different isoforms of the enzyme known as COX-1 and -2. Site directed mutagenesis studies suggest that non-selective COX inhibitors of diverse chemical families exhibit differential binding modes to the two isozymes. These results cannot clearly be explained from the sole analysis of the crystal structures of COX available from X-ray diffraction studies. With the aim to elucidate the structural features governing the differential inhibitory binding behavior of these inhibitors, molecular modeling studies were undertaken to generate atomic models compatible with the experimental data available. Accordingly, docking of different COX inhibitors, including selective and non-selective ligands: rofecoxib, ketoprofen, suprofen, carprofen, zomepirac, indomethacin, diclofenac and meclofenamic acid were undertaken using the AMBER program. The results of the present study provide new insights into a better understanding of the differential binding mode of diverse families of COX inhibitors, and are expected to contribute to the design of new selective compounds.     

多维数据判别分析的非参核密度算法研究

判别分析在数据挖掘、识别中有着广泛的应用,其中充分利用训练集的信息,改进判别规则算法,降低误判率一直是众多研究关注的焦点。传统的一些判别算法中,往往事先假定数据的分布类型来建立判别规则,但多维数据结构往往存在违背假定的情形,从而导致较高的误判率。针对此类问题,提出采用非参核密度算法建立多维数据的判别规则,同时通过Iris数据和Seeds数据进行实证分析。结果表明,与现有的判别分析算法相比较,所提判别算法利用样本资料信息更充分,显著提高了多维数据的判别精度,并且该算法不受分布假定的限制,具有广泛的适用性。     

Development of efficient docking strategies and structure-activity relationship study of the c-Met type II inhibitors

c-Met is a transmembrane receptor tyrosine kinase and an important therapeutic target for anticancer drugs. In the present study, we systematically investigated the influence of a range of parameters on the correlation between experimental and calculated binding free energies of type II c-Met inhibitors. We especially focused on evaluating the impact of different force fields, binding energy calculation methods, docking protocols, conformation sampling strategies, and conformations of the binding site captured in several crystallographic structures. Our results suggest that the force fields, the protein flexibility, and the selected conformation of the binding site substantially influence the correlation coefficient, while the sampling strategies and ensemble docking only mildly affect the prediction accuracy. Structure-activity relationship study suggests that the structural determinants to the high binding affinity of the type II inhibitors originate from its overall linear shape, hydrophobicity, and two conserved hydrogen bonds. Results from this study will form the basis for establishing an efficient computational docking approach for c-Met type II inhibitors design.     

Influence functions for a linear subspace method

A linear subspace method, which is one of discriminant methods, was proposed as a pattern recognition method and was studied. Because the method and its extensions do not encounter the situation of singular covariance matrix, we need not consider extensions such as generalized ridge discrimination, even when treating a high dimensional and sparse dataset. In addition, classifiers based on a multi-class discrimination method can function faster because of the simple decision procedure. Therefore, they have been widely used for face and speech recognition. However, it seems that sufficient studies have not been conducted about the statistical assessment of training data performance for classifier in terms of prediction accuracy. In statistics, influence functions for statistical discriminant analysis were derived and the assessments for analysis result were performed. These studies indicate that influence functions are useful for detecting large influential observations for analysis results by using discrimination methods and they contribute to enhancing the performance of a target classifier.     

Reduction of processing time for optimal and quadratic discriminant analyses

A fast algorithm is presented for optimal discriminant analysis and quadratic discriminant analysis. In this algorithm, the discriminant function of an input feature vector for each category is calculated via a monotonically increasing sequence, and when the sequence value exceeds a certain value, then you can assert that the current category cannot be the classification result and omit the redundant calculation of the remaining terms for the category, thus making the calculation faster. Applying this algorithm to the recognition experiment on handwritten characters, we could reduce the processing time to 4% of the conventional simple method. Since both discriminant analyses assume the normal distribution of the features, disnormality contained in real-world data affects the accuracy of the two discriminant analyses. We also compared the accuracy performances of the two discriminant analyses using real-world data and artificial data.     

DEA-Based Piecewise Linear Discriminant Analysis

Nonlinear classification models have better classification performance than the linear classifiers. However, for many nonlinear classification problems, piecewise-linear discriminant functions can approximate nonlinear discriminant functions. In this study, we combine the algorithm of data envelopment analysis (DEA) with classification information, and propose a novel DEA-based classifier to construct a piecewise-linear discriminant function, in this classifier, the nonnegative conditions of DEA model are loosed and class information is added; Finally, experiments are performed using a UCI data set to demonstrate the accuracy and efficiency of the proposed model.     

Benchmarking binary classification models on data sets with different degrees of imbalance

In practice, there are many binary classification problems, such as credit risk assessment, medical testing for determining if a patient has a certain disease or not, etc. However, different problems have different characteristics that may lead to different difficulties of the problem. One important characteristic is the degree of imbalance of two classes in data sets. For data sets with different degrees of imbalance, are the commonly used binary classification methods still feasible? In this study, various binary classification models, including traditional statistical methods and newly emerged methods from artificial intelligence, such as linear regression, discriminant analysis, decision tree, neural network, support vector machines, etc., are reviewed, and their performance in terms of the measure of classification accuracy and area under Receiver Operating Characteristic (ROC) curve are tested and compared on fourteen data sets with different imbalance degrees. The results help to select the appropriate methods for problems with different degrees of imbalance.     

Molecular docking and analysis of interactions between vascular endothelial growth factor (VEGF) and SPARC protein

The extracellular module of SPARC/osteonectin binds to vascular endothelial growth factor (VEGF) and inhibits VEGF-stimulated proliferation of endothelial cells. In an attempt to identify the binding site for SPARC on VEGF, we hypothesized that this binding site could overlap at least partially the binding site of VEGF receptor 1 (VEGFR-1), as SPARC acts by preventing VEGF-induced phosphorylation of VEGFR-1. To this end, a docking simulation was carried out using a predictive docking tool to obtain modeled structures of the VEGF-SPARC complex. The predicted structure of VEGF-SPARC complex indicates that the extracellular domain of SPARC interacts with the VEGFR-1 binding site of VEGF, and is consistent with known biochemical data. Following molecular dynamics refinement, side-chain interactions at the protein interface were identified that were predicted to contribute substantially to the free energy of binding. These provide a detailed prediction of key amino acid side-chain interactions at the protein-protein interface. To validate the model further, the identified interactions will be used for designing mutagenesis studies to investigate their effect on binding activity. This model of the VEGF-SPARC complex should provide a basis for future studies aimed at identifying inhibitors of VEGF-induced angiogenesis.     

基于GAN-LSTM的APT攻击检测

高级持续性威胁(Advanced Persistent Threat,APT)带来的危害日趋严重。传统的APT检测方法针对的攻击模式比较单一,处理的APT攻击的时间跨度相对较短,没有完全体现出APT攻击的时间序列性,因此当攻击数据样本较少、攻击持续时间较长时准确率很低。为了解决这个问题,文中提出了基于生成式对抗网络(Generative Adversarial Netwokrs,GAN)和长短期记忆网络(Long Short-term Memory,LSTM)的APT攻击检测方法。一方面,基于GAN模拟生成攻击数据,为判别模型生成大量攻击样本,从而提升模型的准确率;另一方面,基于LSTM模型的记忆单元和门结构保证了APT攻击序列中存在相关性且时间间距较大的序列片段之间的特征记忆。利用Keras开源框架进行模型的构建与训练,以准确率、误报率、ROC曲线等技术指标,对攻击数据生成和APT攻击序列检测分别进行对比实验分析。通过生成式模型生成模拟攻击数据进而优化判别式模型,使得原有判别模型的准确率提升了2.84%,与基于循环神经网络(Recurrent Neural Network,RNN)的APT攻击序列检测方法相比,文中方法在检测准确率上提高了0.99个百分点。实验结果充分说明了基于GAN-LSTM的APT攻击检测算法可以通过引入生成式模型来提升样本容量,从而提高判别模型的准确率并减少误报率;同时,相较于其他时序结构,利用LSTM模型检测APT攻击序列有更好的准确率和更低的误报率,从而验证了所提方法的可行性和有效性。     

Combining docking, scoring and molecular field analyses to probe influenza neuraminidase-ligand interactions

In this project, several docking conditions, scoring functions and corresponding protein-aligned molecular field analysis (CoMFA) models were evaluated for a diverse set of neuraminidase (NA) inhibitors. To this end, a group of inhibitors were docked into the active site of NA. The docked structures were utilized to construct a corresponding protein-aligned CoMFA models by employing probe-based (H+, OH, CH3) energy grids and genetic partial least squares (G/PLS) statistical analysis. A total of 16 different docking configurations were evaluated, of which some succeeded in producing self-consistent and predictive CoMFA models. However, the best model coincided with docking the ionized ligands into the hydrated form of the binding site via PLP1 scoring function (r2LOO=0.735, r2PRESS against 24 test compounds=0.828). The highest-ranking CoMFA models were employed to probe NA-ligand interactions. Further validation by comparison with a co-crystallized ligand-NA crystallographic structure was performed. This combination of docking/scoring/CoMFA modeling provided interesting insights into the binding of different NA inhibitors.     

Conformational analysis and docking study of potent factor XIIIa inhibitors having a cyclopropenone ring

A conformational analysis and docking study of potent factor XIIIa inhibitors having a cyclopropenone ring were carried out in an attempt to obtain structural insight into the inhibition mechanism. First, stable conformers of the inhibitors alone were obtained from the conformational analysis by systematic search and molecular dynamics. Next, a binding form model of factor XIIIa was built based on an X-ray crystal structure of the enzyme. Finally, the docking study of the inhibitors into the model’s binding site was performed. From the resulting stable complex structures, it was found that the cyclopropenone ring fits the active site located at the base of the binding cavity with high complementarity. The carbonyl oxygen of the cyclopropenone ring formed a hydrogen bond to the indole NH group of Trp279 and the terminal carbon atom of the reactive C=C double bond was in close proximity to the sulfur atom of the catalytic residue, Cys314. This binding mode suggests a possible inhibition mechanism, whereby the cysteine residue reacts with the cyclopropenone ring of the inhibitor, forming an enzyme-ligand adduct. In addition, the higher interaction energies between factor XIIIa and the inhibitors alluded to the probable binding sites of the ligand side chain.     

A molecular dynamics study of the BACE1 conformational change from Apo to closed form induced by hydroxyethylamine derived compounds

BACE1 is an aspartyl protease which is a therapeutic target for Alzheimer’s disease (AD) because of its participation in the rate-limiting step in the production of Aβ-peptide, the accumulation of which produces senile plaques and, in turn, the neurodegenerative effects associated with AD. The active site of this protease is composed in part by two aspartic residues (Asp93 and Asp289). Additionally, the catalytic site has been found to be covered by an antiparallel hairpin loop called the flap. The dynamics of this flap are fundamental to the catalytic function of the enzyme. When BACE1 is inactive (Apo), the flap adopts an open conformation, allowing a substrate or inhibitor to access the active site. Subsequent interaction with the ligand induces flap closure and the stabilization of the macromolecular complex. Further, the protonation state of the aspartic dyad is affected by the chemical nature of the species entering the active site, so that appropriate selection of protonation states for the ligand and the catalytic residues will permit the elucidation of the inhibitory pathway for BACE1. In the present study, comparative analysis of different combinations of protonation states for the BACE1-hydroxyethylamine (HEA) system is reported. HEAs are potent inhibitors of BACE1 with favorable pharmacological and kinetic properties, as well as oral bioavailability. The results of Molecular Dynamics (MD) simulations and population density calculations using 8 different parameters demonstrate that the LnAsp289 configuration (HEA with a neutral amine and the Asp289 residue protonated) is the only one which permits the expected conformational change in BACE1, from apo to closed form, after flap closure. Additionally, differences in their capacities to establish and maintain interactions with residues such as Asp93, Gly95, Thr133, Asp289, Gly291, and Asn294 during this step allow differentiation among the inhibitory activities of the HEAs. The results and methodology here reported will serve to elucidate the inhibitory pathway of other families of compounds that act as BACE1 inhibitors, as well as the design of better leader compounds for the treatment of AD.