摘要
全球20种最畅销的12种药物都是以GPCR(G蛋白偶联受体)为作用靶标,因为生化实验方法很难得到其三维结构,已知结构的GPCR数目只有7个.当前主流的GPCR结构预测方法主要是基于模板的预测和从头预测,但是由于GPCR序列相似性较低,而且长度普遍较长,故单靠一种方法很难预测出比较好的GPCR三维结构.本文结合GPCR的特点,将上述两种方法进行结合,提出了一种并行化的结构预测方法.使用该方法对GPCR Dock 2010中的两个目标(CXCR4和D3)进行预测,实验结果表明,本文预测的跨膜螺旋区域和ECL2区域比官网发布的大多数研究小组的结果更接近天然结构.
Abstract
GPCRs are the targets of 12 top 20 best-selling drugs in the world. However, the number of known structures of GPCRs is only 7 because it is difficult to get their three-dimensional structure. Two major GPCR structure prediction methods are template-based prediction and ab initio prediction. Due to low sequence similarity and long sequence length, it is very hard to predict three-dimensional structure of GPCR accurately with one of methods. This paper combines the two methods referred to as the parallel method. The results show that the parallel method predicts more native structure than other methods on TM and ECL2 regions.
NETL
NSTL
维普
万方数据