一种基于四层推理模型的旋转异构体库构造方法  

A method for constructing rotamer library based on four level model

作　　者：缪大俊[1] 吕强[1] 黄旭[1] 

机构地区：[1]苏州大学计算机科学与技术学院,江苏苏州215006

出　　处：《苏州大学学报（自然科学版）》2011年第4期29-33,共5页Journal of Soochow University(Natural Science Edition)

基　　金：国家自然科学基金(60970055)

摘　　要：GPCR是一类重要的药物靶标,其侧链构象直接影响着对药物绑定位点的研究.现有的侧链旋转异构体库没有考虑侧链扭角之间的相互影响,而将它们同等看待.本文提出了一种基于四层推理模型的旋转异构体库构造方法,将侧链之间的相互影响也考虑进来,生成了一种富含层次化信息的旋转异构体库.并采用侧链预测程序Rosetta对GPCR Dock 2008和GPCR Dock 2010中的3个目标在本文产生的旋转异构体库的基础上进行预测,其预测精度接近于基于经典的旋转异构体库的侧链预测结果.通过实验还发现,本文提出的采用分层思想的模型生成的旋转异构体库明显优于不采用分层思想的模型生成的旋转异构体库.GPCRs are one of the most important class of drug targets.Their side-chain confirmation have a directly influence on the research of the binding-site of drugs.However,the current rotamer libraries do not consider the interactions between torsion angles of side-chains.On the other hand,they are treated equally.This article presents a method to construct a rotamer library based on a four level model.Those interactions between torsion angles are also taken into consideration,producing a rotamer library which is rich in hierarchical information.Test of 3 targets of GPCR Dock 2008 and GPCR Dock 2010 based on our rotamer library is conducted on Rosetta.By comparison,the side-chain prediction accuracy based on our library is close to the state-of-the-art rotamer library.The paper also finds that the hierarchical model is much better than none hierarchical model through experiments.

关 键 词：GPCR蛋白质 侧链预测 四层推理模型 隐马尔科夫模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]