藉由同源模擬法預測蛋白質三級結構

Abstract

同源模擬法（Homology Modeling）是目前預測蛋白質三級結構成功率最高的方法。同源模擬法主要可分為以下幾個流程：（1）資料庫搜尋及模版選擇；（2）多重序列排列；（3）結構模擬；（4）結構合理性評估。而在模版選擇的部分，一般簡單的序列比對不易搜尋到蛋白質間較遙遠的同源關係。目前，蛋白質序列側寫（profile）比對已經能成功的偵測到蛋白質間較遙遠的同源關係。而我們實驗室也開發了類似的軟體Prof2。在這篇論文中，我們比較了一些模版選擇的方法，蛋白質序列側寫比對的確能偵測到蛋白質較遙遠的同源關係，因此我們利用這個方式進行資料庫搜尋及模版選擇，最後再利用MODELLER產生被預測的蛋白質三級結構。(PS)2是結合以上方法，可自動化的預測蛋白質三級結構。在CASP5及CAFASP3的比較結果中，(PS)2所預測的蛋白質三級結構，與其他自動化的預測方法相比，結果極為相近。

Homology or comparative modeling is currently the most accurate method to predict the three-dimensional (3D) structure of proteins. The prediction process consists of fold assignment, target-template alignment, model building, and model evaluation. In fold assignment, it is difficult to detect remote homologies by basic protein sequence-sequence comparison. Currently, profile-profile comparison methods successfully detect weak similarities between two protein families. Our laboratory also developed the Prof2 profile-profile comparison software. In this study, we compare some protein fold recognition methods. Prof2 has the best performances for fold recognition, and we use this program to select the modeling template(s). The final three-dimensional structure is build using the modeling package MODELLER. We present here (PS)2, an automated homology modeling server which using profile-profile comparison method. The method gets benefit of the modeling templates selection. To evaluate the performances of our modeling server, we test on 17 targets in the CASP5 experiment (Critical Assessment of Techniques for Proteins Structural Prediction) by LGA package. The accuracy of our modeling strategy obtains very similar results, which compared with the automated structure prediction servers in CAFASP3. (PS)2 is available through its web site at http://ps2.life.nctu.edu.tw.