兰州大学计算机化学研究所介绍_23页.ppt

单击此处编辑母版标题样式,单击此处编辑母版文本样式,第二级,第三级,第四级,第五级,*,兰州大学计算机化学研究所介绍,兰州大学计算机化学研究所,Lanzhou University,20,世纪,80,年代,-90,年代中期,，,从分析化学应用开始，率先开展将计算机技术应用于化学的研究，在化学计量学领域开展工作，如在滴定分析的理论计算、分析测试专家系统、,HPLC,中溶质保留行为的研究等方面取得了一系列的成果。,90,年代中期,，,开始将人工神经网络等方法应用在定量结构活性,/,性质关系,(QSAR/QSPR),方面的研究,在开发新的分子描述符,建立结构参数与溶解度、色谱保留值、化合物的热力学性质的关系等方面开展了一系列研究工作。从,1997,年开始，与巴黎第七大学,ITODYS,研究所范波涛教授合作开展计算机化学的研究。,2000,年以来,，,开始将,QSAR/QSPR,的研究与化学数据挖掘技术相结合，开展支撑向量机（,SVM,）及其他核函数方法在化学数据挖掘领域的应用研究；另一方面开始涉及基于结构的药物设计及基于量子化学方法的化学反应机理研究。,研究所简介,(1),研究所简介,(2),2001,成立了兰州大学计算机化学研究所。,最近,5,年来，在,SCI,期刊上发表研究论文,40,余篇，逐步形成兰州大学化学学科中一个新兴交叉学科,-,计算机化学。,2004,年，,已申报一级学科博士学位授予单位自主设立二级学科博士点：,化学与生物信息学。科研梯队：,目前研究人员主由兰州大学化学化工学院、兰州大学信息科学与工程学院的教师和研究生组成，其中教授,5,人，副教授,3,人，讲师,2,人。,主要研究方向：,QSAR/QSPR,化学数据挖掘方法及应用,化学反应机理的研究化学软件设计与开发,药物设计,研究所简介,(3),研究所具备的软硬件设施：,SGI,工作站两台，,ALPHA Server 1,台，,PC,服务器,4,台，,Sybyl6.9,，,Gauss98,SVM,、,ANN,程序包，以及一些分子模型软件等。,可以为“计算化学虚拟实验室”贡献的资源：,从目前的情况看，直接可以供大家使用的软件资源非常有限。但我们有计算机应用技术专业的研究生队伍，承担过通信行业大型软件的设计及开发工作，如果需要，可以参与,“计算化学虚拟实验室”,自主开发相关软件的工作。,需要“计算化学虚拟实验室”和超级计算中心协助解决的问题：,目前我们的研究工作主要是根据自己已有的条件开展的，还提不出十分具体的要求。,未来实验室建设方向,：希望在对计算资源和大型软件包的共享和利用、组织承担国家大型或重点研究项目，组织开发自主知识产权的计算化学专业软件等方面，发挥重要作用。,支撑向量机方法的应用研究,支撑向量机,支撑向量机,（,Support Vector Machine,SVM,）是由,Vapnik及其合作者发明的，在,1992,计算学习理论的会议上，介绍进入机器学习领域，之后受到广泛关注。,90,年代后期得到全面的发展，现已成为机器学习和数据挖掘领域的标准工具。,它是一种,新的机器学习技术,，是建立在统计学习理论的,VC(Vapnik-Chervonenkis,),维理论和结构风险最小化原则,基础上的，根据有限的样本信息在模型的复杂性（即对特定训练样本的学习精度，,Accuracy,）和学习能力（即无错误的识别任意样本的能力）之间寻找最佳折中，以期获得最好的泛化能力,(Generalization Ability),。,支撑向量机方法的优点,它是,专门针对有限样本情况的,，其目标是得到有限信息下的最优解而不仅仅是样本数趋于无穷大时的最优值；,算法最终将转化成为一个二次寻优问题，从理论上说，得到的将是,全局最优,点，解决了在神经网络等方法无法避免的局部最优问题；,算法将实际问题通过,Kernel,函数的非线性变换转换到高维特征空间（,Feature Space,），在高维空间中构造线性判别函数来实现原有空间中的非线性判别函数，有较好的泛化能力，同时它巧妙地解决了维数问题，其,算法复杂度与样本维数无关,。,Study of the Quantitative Structure-Mobility Relationship of Carboxylic Acids in capillary Electrophoresis Based on Support Vector Machines,The selected structural descriptors and linear model,Study of the Quantitative Structure-Mobility Relationship of Carboxylic Acids in capillary Electrophoresis Based on Support Vector Machines,The nonlinear model based on SVM,Quantitative Prediction of,log,k,of Peptides in High-Performance LiquidChromatography Based on Molecular Descriptors by Using the Heuristic Method and Support Vector Machine,The selected structural descriptors and linear model,J.Chem.Inf.,Comput,.,Sci,.,，,2004,，,in press,Quantitative Prediction of,log,k,of Peptides in High-Performance LiquidChromatography Based on Molecular Descriptors by Using the Heuristic Method and Support Vector Machine,The nonlinear model based on SVM,Study of Quantitative Structure-Mobility Relationship of Amino Acids and Peptides in Capillary Electrophoresis Based on Support Vector Machines,Descriptor,Chemical meaning,Coefficient,Std.Error,Beta,t,-test,sig,Intercept,Intercept,40.236,11.093,3.627,0.001,Complementary Information content(order 2),0.105,0.055,0.526,1.891,0.005,Average Bonding Information content(order 2),29.469,10.305,0.561,2.860,0.006,Moment of inertia C,75.450,24.918,0.321,3.028,0.004,Molecular surface area,-0.039,0.012,-0.633,-3.278,0.002,Max partial charge for a O atom,Zefirovs,PC,254.407,69.947,0.239,3.637,0.001,Min partial charge for a O atom,Zefirovs,PC,64.905,26.748,0.165,2.427,0.019,a,R,=0.894;Standard.Error of the estimate=1.882;,RMS,=,1.758;n=55;,F,=31.835.,The selected structural descriptors and linear model,Study of Quantitative Structure-Mobility Relationship of Amino Acids and Peptides in Capillary Electrophoresis Based on Support Vector Machines,The nonlinear model based on SVM,Prediction of,electrophoretic,mobility of substituted aromatic acids in different aqueousalcoholic solvents by capillary zone electrophoresis based on support vector machine,Prediction of,Isoelectric,Point of Amino Acid Based on GA-PLS and,SVMs,The seven descriptors selected using GA-PLS,The optimal QSPR model developed was based on support vector machines,which showed the following results:the root-mean-square error of 0.2383 and the prediction correlation coefficient R)0.9702 were obtained for the whole data set.,An Accurate QSPR Study of O-H Bond Dissociation Energy in Substituted Phenols Based on Support Vector Machines,Support Vector Machines-Based Quantitative Structure-Property Relationship for thePrediction of Heat Capacity,QSAR Models for the Prediction of Binding Affinities to Human Serum Albumin Using the Heuristic Method and a Support Vector Machine,J.Chem.Inf.,Comput,.,Sci,.,2004,in press,QSAR Study of Ethyl 2-(3-Methyl-2,5-dioxo(3-pyrrolinyl)amino-4-(trifluoromethyl)pyrimidine-5-carboxylate:An Inhibitor of AP-1 and NF-,K,B Mediated Gene Expression Based on Support Vector Machines,QSAR and Classification models of a novel series of COX-2 selective inhibitors:1,5-Diarylimidazoles based on support vector machines,Quantitative modelling results in a nonlinear,seven-descriptor model based on,SVMs,with root mean-square errors,0.107,0.136,for training and prediction sets,respectively.,The best classification results are found using,SVMs,:the accuracy for training and prediction sets are,91.2%,for the test set 88.2%,respectively.,Journal of Computer-Aided Molecular Design,Accepted,Diagnosing Breast Cancer Based on Support Vector Machines,Diagnosing Anorexia Based on Partial Least Squares,Back Propagation Neural Network,and Support Vector Machines,Thank you,