基于“实验演化-关联变异”的蛋白结构预测研究

蛋白结构预测方法具有十分重要的意义。近年来通过同源序列中共进化的残基信息预测蛋白分子三维结构的方法越来越受到人们的重视。然而，自然界中许多蛋白的同源序列数目较少；更为关键的是，对蛋白功能有害的突变被自然选择所淘汰，导致对蛋白折叠有负面影响的结构信息不能在现存蛋白序列中体现。此两点大大限制了该方法的应用。.据此，本项目拟建立基于“实验演化-关联变异”的蛋白结构预测方法，利用酵母筛选系统，结合高通量测序技术，快速得到大量的“同源”序列，根据共进化原理提取出接触残基对和不可接触残基对信息，预测蛋白结构。相较于现有共进化预测蛋白结构的方法，该方法能够轻易获得大量的“同源”序列，以及包含有害突变的“同源”序列，使得基于共进化的结构预测更为准确，从而能够十分有效地推广到其它蛋白的结构预测。

Protein structural prediction is very important for complementing the experimental structural biology. Recently, it is attractive to predict protein structures from evolutionary sequence variations. However, there are two limitations by this method: firstly, there are not enough homologous sequences of many proteins; secondly, the harmful structural informations are not included because homologous sequences are the successful results of natural selection. .In this project, a huge number of “homologous” sequences will be obtained by screening experiments in yeast and high throughput sequencing. Amino acid covariations will be extracted from these sequences and used to fold the protein. We try to benchmark the approach that predicting the structure of proteins by experimental evolution, high throughput sequencing and covariation analysis. The approach is more convenient, more accurate and easier to be promoted.