局部結構比對於蛋白質功能性片段的尋找與應用

Abstract

蛋白質的功能與胺基酸鍊的摺疊結構有密切的關係，因此對於蛋白質結構的研究，有助於我們對蛋白質功能的瞭解。而蛋白質結構比對演算法就是其中一種很重要的蛋白質結構研究工具。 本論文提出一種蛋白質局部結構比對的方法，目的是想要找出具有相同生化功能的蛋白質，它們所共有的局部結構片段；之後更進一步利用這些找出來的片段，來做蛋白質功能的預測。整個演算法我們可以分為四大步驟，一、以pair-wise的方式找出蛋白質相似的局部結構片段。二、找出局部結構片段之間的相似性。三、局部結構片段的分群。四、找出代表性的pattern。 在實驗方面，針對已知的蛋白脢，我們將找出來的片段與具有生化功能的基質用jmol3D蛋白質檢視工具顯示出來，可以發現我們所找出來的片段都會在基質的附近，這表示我們所找出來的片段，確實具有某程度的生物意義。另外，將我們的演算法與林育星 95年的「利用蛋白質序列與結構關係預測酵素種類」演算法做比較。我們也確實能補足他的演算法在某些EC number中無法找出pattern的缺陷。

The functions of proteins are mainly affected by their structures. Therefore, the study of protein structures is helpful to realize the protein function. Protein structure comparison algorithm is one of the important tools in protein structure research. In this thesis, we introduce a protein local structure comparison method. We try to find the common local substructures from proteins which have the same biochemical function. And then we can use those substructures to predict the protein function. In our algorithm has four main steps：1. pair-wise protein local structure comparison to find the common local substructures. 2. calculate the similarity between substructures. 3. cluster substructures by similarity 4. find the representative pattern. In our experiment, we use the protein 3D viewer Jmol to show the location of patterns which we find and substrate. We can find out that patterns are near the substrates. Therefore, we can announce the patterns have some biochemical meaning. In addition, we compared our algorithm with 「Enzyme class prediction via mining conserved region in sequence and structure」algorithm（Lin ,2006）. We can find patterns from the EC numbers which it can’t find any patterns.