一种基于SVR的分辨近天然G蛋白耦联受体—配体构象的方法  

A SVR-Based Method for Indentifying Near-Native GPCR-Ligand Confromation Decoys

作　　者：杨凌云[1] 吕强[1,2] 

机构地区：[1]苏州大学计算机科学与技术学院,苏州215006 [2]江苏省计算机信息处理技术重点实验室,苏州215006

出　　处：《生物信息学》2011年第2期167-170,共4页Chinese Journal of Bioinformatics

基　　金：国家自然科学基金(批准号:60970055)资助项目

摘　　要：蛋白质小分子对接的难点之一是从生成的大量候选结构中挑选出近天然构象。本文使用了一种基于SVR的方法来挑选RosettaLigand生成的GPCR—配体decoy构象中的近天然构象。首先,对已有数据训练得到一个SVR模型,预测decoy构象的LRMSD,然后依此挑选近天然构象。最终,比较了本文方法和RosettaLigand方法挑选出的近天然构象decoy的质量,结果优于RosettaLigand方法,结果表明了本文方法能够有效地挑选出近天然构象。Identify the near-native conformation decoys generated by docking algorithm is one difficulty of protein small molecular docking.The paper presents a new method based on SVR to identify the near-native GPCR-Ligand conformation decoys generated by RosettaLigand.We predict the deocy＇s Ligand RMSD using this SVR model and pick the decoys with lowest predicted Ligand RMSD as the near native conformation decoys.At last,we compare the quality of near-native conformation decoys identified by our method and RosettaLigand method,and our method is better than RosettaLigand method.The results shows that our method would efficiently identify the near-native conformation decoys.

关 键 词：GPCR SVR 分子对接 

分 类 号：Q518.2[生物学—生物化学]