基于分子描述符和机器学习方法预测和虚拟筛选MMP-13对MMP-1的选择性抑制剂

基质金属蛋白酶-13(MMP-13)为预防和治疗骨关节炎(OA)提供了充满希望的靶标.通过抑制剂来阻断MMP-13的活性将会对治疗OA疾病产生潜在的作用.然而,宽谱抑制剂同样抑制MMP家族的其它成员,特别是MMP-1,这将会导致肌与骨的综合症.因此,设计和发现潜在的MMP-13相对于MMP-1的高效选择性抑制剂,在对治疗OA新型药物的研发中具有相当重要的现实意义.本研究通过两种机器学习方法(ML):支持向量机(SVM)和随机森林(RF)来建立分类模型,用于预测不同结构的MMP-13对MMP-1的选择性抑制剂.所建这些模型的预测效果都已经达到了令人满意的精度.在这两种ML模型中,RF对于MMP-13选择性抑制剂和非抑制剂的精度分别达到97.58%和100%.同时,与MMP-13对MMP-1的选择性抑制最相关的分子描述符也基于不同的特征选择方法被两种模型挑选出来.最后,用预测效果最好的RF模型虚拟筛选了ZINC数据库的"fragment-like"子集,从而得到了一系列潜在的候选药物.研究表明,机器学习方法,特别是RF方法,对于发现潜在的MMP-13选择性抑制剂十分有效.同时还得到了一些与MMP-13的选择性抑制相关的分子描述符.     

基于分子描述符和机器学习方法预测和虚拟筛选乳腺癌靶向蛋白HEC1抑制剂

HEC1(癌症高表达蛋白)是纺锤体检查点控制、着丝粒功能、细胞存活的关键的有丝分裂调节器,与原发性乳腺癌的不良预后有关.筛选具有高亲和力的HEC1新型抑制剂对探索乳腺癌的靶向治疗具有重要意义.本文从结构多样性的化合物库中筛选HEC1抑制剂.通过对分子描述符的特征筛选,采用支持向量机(SVM)和随机森林(RF)方法分别对HEC1抑制剂和非抑制剂建立了分类模型.经对比, RF模型显示了更好的预测精度.我们采用RF模型对HEC1抑制剂进行了虚拟筛选,从“in-house”实体库筛选得到2个潜在的HEC1抑制剂分子.随后对筛出的化合物进行了体外活性实验,发现对乳腺癌细胞株MDA-MB-468和MDA-MB-231均有一定程度的抗肿瘤活性.研究结果表明,机器学习方法对于设计和虚拟筛选HEC1抑制剂有良好的效果.     

基于机器学习方法预测CDK4的抑制剂

最近的研究已经证实,以周期蛋白依赖性激酶4(cyclin-dependent kinase 4,CDK4)为靶点,通过CDK4的抑制剂重新建立细胞周期的调控在肿瘤靶向治疗的发展中已经成为有吸引力的方向。本文测试了三种机器学习方法,k最近邻、C4.5决策树和随机森林(random forest,RF),用于预测CDK4的抑制剂。所建这些模型都达到了令人满意的预测效果。其中,RF模型当参数Mtry=13、jbt=255时对应的总预测精度最大,为96.65%。同时,与CDK4抑制剂最相关的25个分子描述符也被最优的RF模型挑选了出来。本文的研究表明,机器学习方法特别是RF方法,对于发现潜在的CDK4抑制剂十分有效。     

基于机器学习方法虚拟筛选Syk的抑制剂

脾酪氨酸激酶(Syk)是一种细胞内的酪氨酸激酶细胞质受体,在类风湿性关节炎(RA)的发病过程中起着至关重要的作用。筛选Syk抑制剂对RA的治疗有着重要的意义。采用C4.5决策树与随机森林(RF)两种机器学习方法分别对Syk抑制剂与非抑制剂建立模型,经过对比,RF具有更好的预测精度。采用RF模型对Syk抑制剂进行虚拟筛选,从ZINC数据库筛选得到潜在的Syk抑制剂分子。研究结果表明,机器学习方法对于虚拟筛选和发现潜在的Syk抑制剂十分有效。     

基于机器学习方法的激素敏感脂肪酶抑制剂活性预测(英文)

脂肪组织中,激素敏感脂肪酶(HSL)被认为是调节脂肪酸代谢的关键限速酶.HSL在糖尿病的发病过程中起重要作用,抑制HSL活性有助于糖尿病的治疗,因此探索新颖的HSL抑制剂成为当前研究的热门.在激素敏感脂肪酶的作用机制和三维结构缺乏的情况下,需要发展预测HSL抑制剂的方法.本文采用几种机器学习方法(支持向量机(SVM)、k-最近相邻法(k-NN)和C4.5决策树(C4.5DT))对已知的HSL抑制剂与非抑制剂建立分类预测模型.252个结构多样性化合物(123个HSL抑制剂与129个HSL非抑制剂)被用于测试分类预测系统,并用递归变量消除法选择与HSL抑制剂相关的性质描述符以提高预测精度.本研究对独立验证集的总预测精度为75.0%-80.0%,HSL抑制剂的预测精度为85.7%-90.5%,非HSL抑制剂的预测精度为63.2%-68.4%.支持向量机方法给出最好的总预测精度(80.0%).本研究表明支持向量机等机器学习方法可以有效预测未知数据集中潜在的HSL抑制剂,并有助于发现与其相关的分子描述符.     

基于机器学习方法的H1N1神经氨酸苷酶抑制剂的分类预测

流感是一种主要的呼吸道传染病, 在普通人群中有着较高的发病率, 而对于一些年老和高危病人还有较高的死亡率. 研究显示抑制神经氨酸苷酶(NA)可以阻断病毒RNA复制, 因此NA是有效治疗H1N1型流感病毒的重要药物靶标. 通过计算机方法进行虚拟筛选和预测NA抑制剂已经变得越来越重要. 针对酶活性位点进行基于结构的合理药物设计, 开发H1N1 病毒神经氨酸苷酶抑制剂, 已成为药物研究的热点之一. 本文通过多种机器学习方法(支持向量机(SVM)、k-最近相邻法(k-NN)和C4.5决策树(C4.5DT))对已知的神经氨酸苷酶抑制剂(NAIs)与非神经氨酸苷酶抑制剂(non-NAIs)建立分类预测模型. 其中227个结构多样性化合物(72个NAIs与155个non-NAIs)被用于测试分类预测系统, 并用递归变量消除法选择与神经氨酸苷酶抑制剂分类相关的性质描述符以提高预测精度. 本研究对独立验证集的总预测精度为75.9%-92.6%, NA 抑制剂的预测精度为64.3%-78.6%, 非H1N1抑制剂的预测精度为77.5%-97.5%. SVM法给出最好的总预测精度(92.6%). 本研究表明支持向量机等机器学习方法可以有效预测未知数据集中潜在的NA抑制剂, 并有助于发现与其相关的分子描述符.     

机器学习方法用于选择性环氧化酶-2抑制剂活性预测模型的建立

与传统的非甾体类消炎药相比,选择性环氧化酶-2抑制剂具有无胃肠道粘膜损伤,溃疡和肾功能障碍等严重的副作用,设计选择性环氧化酶-2抑制剂具有重要意义。本文用支持矢量学习机和神经网络两种机器学习方法建立选择性环氧化酶-2抑制剂的活性预测模型,以期为选择性环氧化酶-2抑制剂药物的合成提供先导化合物。我们将467个环氧化酶-2抑制剂用Kennard-Stone方法分为训练集,验证集和独立测试集,对每一抑制剂分子我们计算了463个包含组成描述符和拓扑描述符的分子描述符来表征其分子结构,并通过F-Score方法选取最重要的分子描述符用于分类模型的建立。结果表明,SVM方法通过变量筛选后具有很好的预测能力,其预测正确率达到93.30%。     

基于机器学习方法的丙型肝炎病毒非结构蛋白5B聚合酶抑制剂活性预测

在丙型肝炎病毒(HCV)的基因复制和蛋白质成熟的过程中, 非结构蛋白5B(NS5B)作为RNA依赖的RNA聚合酶起到了重要的作用. 抑制NS5B聚合酶可以阻止丙型肝炎病毒的RNA复制, 因此成为一种治疗丙型肝炎的有效方法. 通过计算机方法进行虚拟筛选和预测NS5B聚合酶抑制剂已经变得越来越重要. 本文主要采用机器学习方法(支持向量机(SVM)、k-最近相邻法(k-NN)和C4.5决策树(C4.5 DT))对已知的丙型肝炎病毒NS5B蛋白酶抑制剂与非抑制剂建立分类预测模型. 1248个结构多样性化合物(552个NS5B抑制剂与696个非NS5B抑制剂)被用于测试分类预测系统, 并用递归变量消除法选择与NS5B抑制剂相关的性质描述符以提高预测精度. 独立验证集的总预测精度为84.1%-85.0%, NS5B抑制剂的预测精度为81.4%-91.7%, 非NS5B抑制剂的预测精度为78.2%-87.2%. 其中支持向量机给出最好的NS5B抑制剂预测精度(91.7%); C4.5决策树给出最好的非NS5B抑制剂预测精度(87.2%); k-最近相邻法给出最好的总预测精度(85.0%). 研究表明机器学习方法可以有效预测未知数据集中潜在的NS5B抑制剂, 并有助于发现与其相关的分子描述符.     

基于分子对接和QSAR方法预测B-Raf II型抑制剂活性

B-Raf激酶在促分裂素原活化蛋白激酶(MAPK)信号转导通路中起着重要作用,已被确定为癌症治疗非常有吸引力的靶标.新型高效B-Raf抑制剂的开发成为癌症治疗的一个热门研究领域.本文以结构多样的B-Raf II型抑制剂为研究对象,联合应用分子对接和定量构效关系(QSAR)模型研究其定量构效关系去探讨抑制活性的起源.两个主题作为研究重点:生物活性构象和描述符.首先对分子对接方法(Glide、Gold、LigandFit、Cdocker和Libdock)进行准确性评价,后将研究的对象分子对接到B-Raf活性位点并获得生物活性构象.基于准确的对接结果,计算得到16个打分评价函数和21个能量描述符,以此构建定量构效关系模型. QSAR结果表明模型具有高度精确的拟合和强的预测能力(模型M1: r² = 0.852, r_(CV)² = 0.790, r_pre² = 0.864;模型M2: r² = 0.738, r_(CV)² = 0.812, r_pre² = 0.8605).同时探讨了对抑制活性有重要影响的描述符,结果表明打分评价函数(G_Score, -ECD, Dock_Score, PMF)与能量描述符(S(hb_ext), DE(int), Emodel)对抑制活性影响非常大.通过虚拟筛选和QSAR模型理论预测,一些新的具有潜在抑制活性的化合物作为B-Raf II型抑制剂被获得.上述信息对于进一步设计新颖高效的B-Raf II型抑制剂提供了有用的指导.     

黄酮类化合物抑制MMP-9的定量结构-活性关系和结构修饰的理论研究

运用密度泛函理论(DFT)和线性回归分析等方法研究了黄酮类化合物抑制MMP-9的定量结构-活性关系(QSAR)和结构修饰. 研究发现, 黄酮类化合物抑制MMP-9的实验生物活性数据－lg EC₅₀值与计算获得的黄酮类化合物的最低空分子轨道能量及分子水合能之间均存在着良好的线性关系|留一法交叉验证结果表明, 所建立的两个相应的QSAR模型都具有良好的稳定性和预测能力. 进一步研究发现, 在黄酮类化合物分子的A环、B环和C环上的合适部位选用供电子能力较强、能降低分子水合能的取代基团对其进行结构修饰, 有利于提高修饰后的分子抑制MMP-9的生物活性. 根据对木犀草素(Luteolin)分子进行结构修饰的结果(共33种化合物), 我们提出了黄酮类化合物抑制MMP-9可能的作用机理, 并设计出8种经结构修饰后生物活性有显著提高的MMP-9抑制剂, 希望将来得到实验的证实.     

Structural basis for the highly selective inhibition of MMP-13

Inhibitors for matrix metalloproteinases (MMPs) are under investigation for the treatment of cancer, arthritis, and cardiovascular disease. Here, we report a class of highly selective MMP-13 inhibitors (pyrimidine dicarboxamides) that exhibit no detectable activity against other MMPs. The high-resolution X-ray structures of three molecules of this series bound to MMP-13 reveal a novel binding mode characterized by the absence of interactions between the inhibitors and the catalytic zinc. The inhibitors bind in the S1' pocket and extend into an additional S1' side pocket, which is unique to MMP-13. We analyze the determinants for selectivity and describe the rational design of improved compounds with low nanomolar affinity.     

Potent, selective pyrone-based inhibitors of stromelysin-1

In an effort to develop alternatives to hydroxamate-based matrix metalloproteinase inhibitors (MPIs), we have utilized the drug discovery program LUDI enhanced with the structural coordinates of a bioinorganic model complex. This method has yielded the first pyrone-based MPIs. The inhibitors demonstrate nanomolar potency against MMP-3 and are selective for MMP-3 over MMP-2 and MMP-1. We postulate that the potency and unusual selectivity profile of these MPI is attributable to the pyrone chelating group.     

Identification of MMP-1 and MMP-9 inhibitors from the roots of Eleutherococcus divaricatus,and the PAMPA test

The purpose of this study was the isolation of metalloproteinases MMP-1 and MMP-9 inhibitors from the chloroform extract of the Eleutherococcus divaricatus roots. Using GC-MS, ¹H and ¹³C NMR, HMQC, HMBC, COSY and DEPT, (+)-sesamin has been identified as a new anti-MMP inhibitor. We report for the first time that (+)-sesamin inhibited MMP-1 and MMP-9 activity in 40% and 17%, respectively. The high inhibitory potential has been shown by ursolic acid (90.9% and 89.8% for MMP-1 and MMP-9). In the PAMPA test, the Pe value for sesamin was established as 17.4 × 10^? 6 cm/s, that for ursolic acid as 30.0 × 10^? 6 cm/s. Verapamil and theophylline were used as a positive and negative control (Pe 42.1 and 2.9 × 10^? 6 cm/s). To our best knowledge, no information was available on this activity of sesamin and other compounds. These studies provide a biochemical basis for the regulation of MMP-1 and MMP-9 by E. divaricatus compounds.     

In silico prediction and screening of γ‐secretase inhibitors by molecular descriptors and machine learning methods

γ‐Secretase inhibitors have been explored for the prevention and treatment of Alzheimer's disease (AD). Methods for prediction and screening of γ‐secretase inhibitors are highly desired for facilitating the design of novel therapeutic agents against AD, especially when incomplete knowledge about the mechanism and three‐dimensional structure of γ‐secretase. We explored two machine learning methods, support vector machine (SVM) and random forest (RF), to develop models for predicting γ‐secretase inhibitors of diverse structures. Quantitative analysis of the receiver operating characteristic (ROC) curve was performed to further examine and optimize the models. Especially, the Youden index (YI) was initially introduced into the ROC curve of RF so as to obtain an optimal threshold of probability for prediction. The developed models were validated by an external testing set with the prediction accuracies of SVM and RF 96.48 and 98.83% for γ‐secretase inhibitors and 98.18 and 99.27% for noninhibitors, respectively. The different feature selection methods were used to extract the physicochemical features most relevant to γ‐secretase inhibition. To the best of our knowledge, the RF model developed in this work is the first model with a broad applicability domain, based on which the virtual screening of γ‐secretase inhibitors against the ZINC database was performed, resulting in 368 potential hit candidates. © 2009 Wiley Periodicals, Inc. J Comput Chem, 2010     

Hydroxamate类抑制剂与MMP-2的分子对接研究

通过分子动力学模拟研究了MMP-2和hydroxamate抑制剂之间的作用模式。在分子动力学模拟中,对于催化区的锌离子和其共价结合的配体(包括抑制剂和组氨酸)采用了键合的模型。从模拟的结果可以看到,R^1取代基和MMP-2的S1疏水口袋中的部分残基能形成很好的几何匹配,从而可以产生很强的范德华和疏水相互作用。模拟结果也表明,两个抑制剂和MMP-2之间分别能形成5个和8个氢键,抑制剂B比A活性更高的原因就是能够形成更加有利氢键作用模式。在整个模拟过程中,催化锌都能保持好的五配位形式,配位键的长度也处于稳定的状态,预测得到的MMP-2和其抑制剂的相互作用模式对于全新抑制剂的设计提供了非常重要的结构信息。     

Computational insights into the selectivity mechanism of APP-IP over matrix metalloproteinases

In this work, selectivity mechanism of APP-IP inhibitor (β-amyloid precursor protein-derived inhibitory peptide) over matrix metalloproteinases (MMPs including MMP-2, MMP-7, MMP-9 and MMP-14) was investigated by molecular modeling methods. Among MMPs, MMP-2 is the most favorable one for APP-IP interacting based on our calculations. The predicted binding affinities can give a good explanation of the activity difference of inhibitor APP-IP. In Comparison with MMP-2/APP-IP complex, the side chain of Tyr214^MMP-7 makes the binding pocket so shallow that the whole side chain of Tyr3^APP-IP can not be fully embraced, thus unfavorable for the N-terminal of APP-IP binding to MMP-7. The poor selectivity of APP-IP toward MMP-9 is mainly related with the decrease of interaction between the APP-IP C-terminal and MMP-9 due to the bulky side chains of Pro193 and Gln199, which is in agreement with experiment. The mutations at residues P193A and Q199G of MMP-9 alternate the binding pattern of the C-terminal of APP-IP by forming two new hydrogen bonds and hydrophobic interactions with MMP-9. The mutants favor the binding affinity of MMP-9 largely. For MMP-14/APP-IP, the large steric effect of Phe204^MMP-14 and the weak contributions of the polar residues Asn231^MMP-14 and Thr190^MMP-14 could explain why MMP-14 is non-selective for APP-IP interacting. Here, the molecular modeling methods were successfully employed to explore the selective inhibitor of MMPs, and our work gives valuable information for future rational design of selective peptide inhibitors toward individual MMP.     

Generation of oxamic acid libraries: antimalarials and inhibitors of Plasmodium falciparum lactate dehydrogenase

Lactate dehydrogenase (LDH) is a key enzyme in the glycolytic pathway of Plasmodium falciparum (pf) and has several unique amino acids, related to other LDHs, at the active site, making it an attractive target for antimalarial agents. Oxamate, a competitive inhibitor, shows high substrate affinity for pfLDH. This class of compounds has been viewed as potential antimalarial agents. Thus, we have developed an effective automated synthetic strategy for the rapid synthesis of oxamic acid and ester libraries to screen for potential lead inhibitors. One hundred sixty-seven oxamic acids were synthesized using a "catch and release" method with overall yields of 20-70%. Most of the compounds synthesized had some inhibitory effects, but compounds 5 and 6 were the most active against both chloroquine- and mefloquine-resistant strains with IC50 values of 15.4 and 9.41 microM and 20.4 and 8.40 microM, respectively. Some oxamic acids showed activities against pfLDH and mammalian LDH (mLDH) at the micromolar range. These oxamic acids selectively inhibited pfLDH 2-5 fold over mLDH. Oxamic acid 21 was the most active against pfLDH at IC50 = 14 and mLDH at IC50 = 25 microM, suggesting that oxamic acid derivatives are potential inhibitors of pfLDH and that further study is required to develop selective inhibitors of pfLDH over mLDH.