搜索
2010 © CloudScientificAll Rights Reserved基于配体的药物设计及在药物研发方面的应用上海康昱盛信息科技有限公司黄韬Tel: 021‐5497 5000‐8003E‐mail: thuang@cloudscientific.com2010 © CloudScientificAll Rights Reservedabout cloudscientific2010 © CloudScientificAll Rights ReservedBest Solutions for Best Research上海康昱盛信息科技有限公司公司介绍3软件•服务2010 © CloudScientificAll Rights Reserved我们致力于为中国的生命科学研发机构提供完整的科学信息解决方案我们于2009年1月从英国帝国理工大学所属的InforSense公司分出，在上海设立总部及办公室。我们由一群经验丰富的行业专家组成，所有人都有生物医药及信息学的背景目前我们已经有100多个客户目前我们的技术服务及咨询团队有10个人，每个人都有硕士以上学位我们与中国的许多顶级科研机构有广泛的联系与合作关于康昱盛2010 © CloudScientificAll Rights Reserved我们能为您提供以下方面的价值•产品:世界领先的生命科学软件产品•服务：数据分析服务、定制开发服务、培训服务2010 © CloudScientificAll Rights Reserved我们代理如下公司软件产品基因组学蛋白组学生物通路网络分析药物设计发现云计算平台2010 © CloudScientificAll Rights Reserved应用领域生物学生物信息学基因组学蛋白组学结构生物学信号转导通路分析药物发现/设计基于大分子结构的基于小分子配体的基于片段的药物开发(临床前及临床试验)药代动力学吸收/分布/排除代谢毒性检测先导化合物优化临床数据管理临床数据统计分析上市转化医学2010 © CloudScientificAll Rights Reservedligand‐based drug design2010 © CloudScientificAll Rights ReservedLigand‐Based Drug Design•Only the active ligand(s), to find a new molecule at least as good as, or better than the original ligand. No receptor structure?2010 © CloudScientificAll Rights ReservedHow to discover or design a new active ligand?•SimilarityIf the new found molecule is similar with the original active ligand, the potency of the new one may be close to the old one’s. Quantitative in similarity, qualitative in potency prediction.•PharmacophorePharmacophoremodel derived from one or more active ligands. Qualitative method to predict ‘active’ or ‘not active’. •QSARQuantitative model derived from a training set of active ligands, could predict the potency value.2010 © CloudScientificAll Rights ReservedFingerprint Model10001010101010…Feature Tree Model3D Alignment ModelPharmacophore Model2D QSAR ModelY=a*d1+b*d2+…DatabaseX‐Modelscreening3D QSAR Model2010 © CloudScientificAll Rights ReservedHow to measure similarity?•Fingerprint•Feature Tree•3D Alignment2010 © CloudScientificAll Rights ReservedFingerprintMol A.Mol BA: 10000111100001100100000100101010010001010B: 10000111100001100100000100101010010110011FingerprintTanimotoCoefficient = #(A B) #(A B) = 0.902010 © CloudScientificAll Rights Reserved3D Alignment (Superpose) Evaluate similarity based on shape, pharmacophore, or electrostatic distribution 2010 © CloudScientificAll Rights ReservedFeature TreeFTreesSimiliarity= 0.81M.Rareyand S.Dixon, JCAMD, 12 (1998)FTreesis developed by BioSolveIT©NNH+NNNH2NH2NONHOOOO2½DFTreesdescriptor:Feature TreeVol.:44.27Ringcl:1Profile:AroDonAccAmdHydDonAccAroAmdHydVol.:33.9Ringcl:0Profile:2010 © CloudScientificAll Rights ReservedSimilarity‐based ClusteringTesting all compounds vs.Testing each compound for each cluster ?Building predicted model for all birds vs. Building predicted models chickens, ducks and gooses? 2010 © CloudScientificAll Rights ReservedSimilarity‐based ClusteringABC…YZA101…01B011…10C111…01…………1……Y010…10Z101…01Jarvis‐Patrick AlgorithmIf A is similar with B, then 1;If not, then 0.Overlap = #(A B) #(A B) ＝0.90If overlap 0.85, then mol A and mol B belong to the same cluster.2010 © CloudScientificAll Rights ReservedSimilarity‐based Diverse Subset SelectionThe farthest molecules in a set (cluster).1stRun2ndRun3rdRunChoosing representative molecules in one set.2010 © CloudScientificAll Rights ReservedPharmacohporeModel Building and Searching2010 © CloudScientificAll Rights ReservedConsensus pharmacophoremodel from 3D alignment results2010 © CloudScientificAll Rights ReservedPharmacophoreElucidation2010 © CloudScientificAll Rights Reserved2D QSARTo establish the relationship between descriptors and activity.Dataset separation is really important before building a QSAR model. We can use PCA, similarity‐based clustering, or pharmacophoremodel to partition the data.r2= 0.95q2=0.752010 © CloudScientificAll Rights Reserved3D QSAR2010 © CloudScientificAll Rights Reserved抗球虫三嗪分子定量构效关系分析及设计中国农业科学院上海兽医研究所薛飞群教授实验室康昱盛信息科技有限公司案例研究2010 © CloudScientificAll Rights Reserved背景知识66 molecules from the original paper: JMC, 1992, 35, 2543‐2550.1. What makes the molecule be ‘active’ or ‘not active’? 2. Can we design new active molecules?2010 © CloudScientificAll Rights Reserved骨架分析MOE SAReport: Detecting the scaffold of all molecules (n=66).Conclusion: They all share the same scaffold.2010 © CloudScientificAll Rights ReservedSAReport建议活性分子1. Enumerating all of the input molecules.2. Performing single‐point mutations at each of the substituent positions3. The unique list of chimericmolecules is then scored according to an estimate of probability2010 © CloudScientificAll Rights Reserved活性药效团分析Analyzed by MOEFlexible Alignment: Top 5 active molecules.Consensus Ph4: share the same pattern R1 is HydR2 is HydR3 contains Aro2010 © CloudScientificAll Rights Reserved将分子按照特征来分类3 bits feature fingerprintR1is Hyd?R2 is Hyd?R3 containsAro?Yes111No000We got 8 patterns.FeatureFingerprintPatterns111R1=Hyd& R2=Hyd& R3=Aro110R1=Hyd& R2=Hyd& R3!=Aro101R1=Hyd& R2!=Hy