BRD4专用虚拟筛选打分函数BRD4LGR:发展及应用
详细信息 查看官网全文
- 英文论文题名：BRD4 Target-specific Scoring Function BRD4LGR:Method Development and Application
- 论文作者：邢婧 ; 谢轶谦 ; 刘荣峰 ; 史喆 ; 刘育志 ; 罗成 ; 郑明月 ; 陈凯先 ; 蒋华良
- 英文论文作者：Jing Xing ; Yiqian Xie ; Rongfeng Liu ; Zhe Shi ; Yuzhi Liu ; Cheng Luo ; Mingyue Zheng ; Kaixian Chen ; Hualiang Jiang ; Shanghai Institute of Materia Medica ; Chinese Academy of Sciences ; Shanghai ChemPartner Co. ; Ltd
- 年：2016
- 作者机构：中国科学院上海药物研究所药物发现与设计中心;上海睿智化学研究有限公司;
- 论文关键词：药物设计 ; 虚拟筛选 ; 专用打分函数 ; 机器学习 ; 结构活性关系
- 英文论文关键词：drag design ; virtual screening ; target-specific scoring function ; machine learning ; structure-activity relationship
- 会议召开时间：2016-07-01
- 会议录名称：中国化学会第30届学术年会摘要集-第二十五分会：化学信息学与化学计量学
- 语种：中文
- 分类号：TQ460.1
- 学会代码：ZGHY
- 会议名称：中国化学会第30届学术年会-第二十五分会 ; 化学信息学与化学计量学
- 会议地点：中国辽宁大连
- 主办单位：中国化学会
- 学会名称：中国化学会
- 页数：1
- 文件大小：231k
- 原文格式：D
- 会议级别：全国
摘要
在新药先导化合物的发现与设计过程中,虚拟筛选具有低成本、高效率的优点,其中分子对接在发现新骨架活性化合物和结构优化中发挥重要作用。然而,目前分子对接程序采用的通用型基于力场打分函数仍普遍具有假阳性率高,靶标体系依赖性强等问题。一个重要的改进方向是发展靶标专用型基于知识的打分函数,比如通过机器学习的方法,用特定药物靶点与小分子结合的结构活性数据建模。在本研究中,我们设计了基于逻辑回归的以描述关键残基与配体相互作用为指标的打分模型。这一模型可以有效区分有活性和非活性化合物,并可用于分析化合物的结构活性关系以及重要分子间相互作用。结合已报道的活性数据和对接-实验筛选的结果,我们将这种计算模型用于发展针对表观遗传学重要药物靶标BRD4的专用打分函数BRD4LGR,筛选富集率较主流通用型打分函数高出20~30%。此外,模型反映的结构活性关系,如小分子与保守水分子或WPF shelf的重要相互作用,可用于进一步指导活性化合物结构改造工作。
In virtual screening(VS),molecular docking contributes a lot in new scaffold hits discovery and hit-to-lead optimization.However,the general purpose scoring functions of docking suffer from problems such as high false-positive rate and target dependency.One of the effective solutions is,to develop target-specific knowledge-based scoring functions.In this research,we designed a logistic regression scoring model based on descriptors of key intermolecular interactions,which can discriminate actives and inactives,also helpful in analyzing main interactions between ligand and the binding pocket.Combining published structure-activity data and our VS results,this computational model was applied to developing BRD4 specific scoring function(BRD4LGR),which has 20-30%higher enrichment rate than traditional ones.The implied SAR,such as crucial interactions between ligand and conserved waters or WPF shelf,may provide guidance for hit-to-lead optimization.
In virtual screening(VS),molecular docking contributes a lot in new scaffold hits discovery and hit-to-lead optimization.However,the general purpose scoring functions of docking suffer from problems such as high false-positive rate and target dependency.One of the effective solutions is,to develop target-specific knowledge-based scoring functions.In this research,we designed a logistic regression scoring model based on descriptors of key intermolecular interactions,which can discriminate actives and inactives,also helpful in analyzing main interactions between ligand and the binding pocket.Combining published structure-activity data and our VS results,this computational model was applied to developing BRD4 specific scoring function(BRD4LGR),which has 20-30%higher enrichment rate than traditional ones.The implied SAR,such as crucial interactions between ligand and conserved waters or WPF shelf,may provide guidance for hit-to-lead optimization.
引文
[1]Shen,Q.;Xiong,B.;Zheng,M.;Luo,X.;Luo,C;Liu,X.;Du,Y.;Li,J.;Zhu,W.;Shen,J.;Jiang,H.J.Chem Inf.Model.2011,51(2):386.
[2]Filippakopoulos,P.;Knapp,S..Nat Rev Drug Discov.2014,13(5):339.
[2]Filippakopoulos,P.;Knapp,S..Nat Rev Drug Discov.2014,13(5):339.