ATP作用區域為基之蛋白質分群與交互作用分析

Abstract

近年來，隨著大規模基因體學與蛋白質體學計畫的發展，人們對生物系統的瞭解也迅速的成長，我們可以透過PDB資料庫，取得愈來愈多被結晶出來的蛋白質立體結構。其中，有許多蛋白質的配體也一併被結晶在結構中。這樣大量的蛋白質-配體結構資訊，使得以結構為基之蛋白質-配體間交互作用分析獲得頗大的助益。然而，一些知名的蛋白質分類資料庫，例如SCOP、CATH等，由於資料庫更新速度過慢，不能跟上解蛋白質結晶結構的速度，當新的蛋白質結晶結構被解出來後，皆無法儘速將之分類，以致影響研究者對蛋白質的結構、功能、配體結合作用力等重要議題上做深入探討。 在本碩士論文研究中，我們發展一套簡單快速的方法論，用以分析蛋白質-配體結構，並且使用ATP結合蛋白作為研究例子。本方法的核心理念乃是根據蛋白質的結構相似度與蛋白質-配體的交互作用側寫，將蛋白質-配體複合物做快速分類。同時也能藉由蛋白質-配體間交互作用的資訊，找出功能性殘基與模版。對於結構相似度，我們同時考慮整個蛋白質或配體結合部位的結構。我們利用快速結構相似度搜尋工具—3D-BLAST，迅速地在整個蛋白質資料庫裡尋找與配體結合蛋白質相似的結構。接著將結合位含有配體的蛋白質結構，以CE做詳細的結構比對，檢查全蛋白與配體作用區域的結構相似性，並將蛋白質做初步分群。對於蛋白質-配體間交互作用側寫，我們則是利用軟體辨認出蛋白質-配體間的交互作用。最後，根據結構相似度及功能性交互作用模版，我們將這套分類蛋白質的方法論應用在ATP結合蛋白質複合物。 分群結束後，我們比較其結果與SCOP資料庫的分類，以每群中佔最多數的SCOP family視為該群的正確答案，且同一SCOP family可同時為多群的答案。在此比較的依據下，結果獲得了95%的正確率。接著，我們系統地對每群中的ATP結合蛋白進行ATP作用區域之交互作用分析，包括氫鍵、π- π疊合作用與正離子-π等三種交互作用，將每一群中交互作用所表現的保守性，建立出各群特有的ATP結合motif。結果發現，我們所找出來的ATP結合模版不但符合目前研究已發現的模版，甚至也另有發現目前資料庫中所沒有定義，可能是新的ATP結合模版。 本論文應用了3D-BLAST，藉由其結構快速搜尋的特性，大幅降低將相似結構分群的時間，並且針對每一群的蛋白質裡找出包含交互作用資訊的ATP結合模版。未來，我們可以利用分群結果及ATP結合模版，來對新結晶或未包含ATP蛋白質作分析與分類。同時也能輕易地將本方法應用於其他重要的蛋白質-配體複合物的研究上。

In recent years, information about biological systems has grown rapidly, in particular through large-scale and global approaches addressing DNA sequence (genomics), protein structure (structural genomics) and protein expression and interactions (proteomics). More and more three-dimensional protein structures have been deposited in the Protein Databank. Many of them are protein-ligand complex structures. This enormous increase in the number of known protein-ligand complexes has therefore had a profound effect on structure-based protein-ligand interaction analyses. However, the classification databases, such as SCOP and CATH, are updated too slowly to classify these rapidly increasing complexes. It is hard to classify newly solved protein structure immediately. In this work, we have developed a very fast method for protein-ligand complexes analysis and used ATP-binding proteins as a study case. The core idea of this method is to cluster protein-ligand complexes based on binding-site structural similarity and protein-ATP interaction profiles. Naturally, this new method is able to analyze the protein-ligand interactions and identify function residues and patterns. For structural similarity, we considered the similarities of both whole proteins and ligand-binding sites. First, we used 3D-BLAST to perform protein-ligand complexes homologous search in whole protein database. Second, CE was used as a detailed structure alignment tool to identify structural similarity of ligand-binding site. Accordingly, we can obtain a preliminary classification for protein complexes. For protein-ligand interaction profiles, the HBPLUS and an in-house software, PiFinder, are used to identify the non-bonded interactions. According to structural similarity and functional protein-ligand interaction patterns, a simple cluster method was applied to group protein-ATP complexes. To evaluate our clustering results, we compared our results to the SCOP classification. The most popular SCOP family in a cluster is set to the representative family of the cluster. Assigning one SCOP family to multiple clusters is also taken as correct answers. Overall, we got a 95% accuracy of the clustering results. We systematically analyzed the non-bonded interactions, including hydrogen bond, π- π stacking, and cation- π interactions, between ATP and the binding protein chains. We found that the three types of non-bonded interactions show relatively strong conservation within clusters. Not only had the ATP-binding motif discovered in the previous works, some novel potential ATP-binding motifs were also identified in some clusters. In this work, 3D-BLAST was applied for fast database search and reducing the time consuming of structure clustering. Furthermore, we can identify ATP-binding motif in each cluster results. In the future, we may use cluster result and ATP-binding motif to analyze and classify new crystal structure. Furthermore, this new method is easily applied to fast analyze other protein-ligand complexes.