利用基因演算法結合二級結構資訊與氨基酸序列比對以達成蛋白質結構相似度之辨識：應用於TIM-barrel結構辨識

Abstract

對於蛋白質的結構比對，序列比對通常是一個標準的取得第一手資訊的方式。但目前的序列比對工具（如BLAST）由於某些演化上的因素而無法找出一些在結構上或功能上相關卻在序列上較無關的蛋白質的同源性。本研究嘗試把序列比對與二級結構的資訊整合成一個新的方法以便於發現更多針對某種蛋白質結構的同源性。首先，經由基因演算法而產生出許多的線性評分函式。並且可以從中得到針對此類結構的相似度門檻。之後藉由投票的機制能得到此蛋白質結構關係的的結論。本研究所建構的方法可以運用在所有的蛋白質結構上。不過，基於TIM-barrel結構的重要性，本研究目前將之運用在TIM-barrel結構辨識上。

For protein structure similarity, the sequence alignment is a standard tool to obtain the first clue. But present sequence alignment tools, such as BLAST, can not detect homologies of distantly relative proteins due to certain evolutionary factors. This thesis developed a proposed method which integrates sequence alignment and secondary structure information to detect more homologies for specified structures. Several linear scoring functions were first constructed by genetic algorithm and some thresholds will be obtained for structural similarity. Though voting mechanism the proposed method can make a resolution of their structural relationship. This method should be able to be applied to all structures. Nevertheless, the thesis only applied it to TIM-barrel structure because of the importance of TIM-barrel.