利用空間相關結構片段快速尋找蛋白質結合片段與環境之研究

Abstract

了解藥物或是化合物其潛在具有結合能力的蛋白質是相當重要地。如在早期藥物發展，即能夠偵測出可能造成的副作用而避免無謂的藥物開發成本與時間；同時，對於已知功能的藥物也能予以提供新的治療用途；並且，目前對於治療具有複雜機制的疾病如癌症及糖尿病, 同時針對疾病中多個標靶蛋白的藥物也是現今藥物發展的一個新策略。所以，尋找可能與化合物結合的潛在標靶蛋白質(potential target proteins)對於學術研究以及臨床藥物開發上都是一大重要議題。 一般相信擁有相似結合環境的蛋白質能與相同或相似結構的化合物產生交互作用。以往研究者常利用比對蛋白質序列或搜尋相似蛋白質結構來尋找潛在的蛋白質，然而過去研究顯示，能結合相同化合物的蛋白質在序列或整體結構上並非都有明顯的相似性及演化關係，而是只在結合環境(binding environment)上相似。因此，針對尋找擁有相似結合環境的蛋白質，我們提出空間相關結構片段(Space-Related Pharmamotif, SRP)的概念，藉由SRP 來搜尋擁有相似結合環境的潛在蛋白質，進一步瞭解蛋白質－化合物交互作用與結合環境的關係。 SRP 是由一組鄰近蛋白質－小分子結合位(binding site)、且三度空間中不連續的蛋白質結構片段所組成，相較於以往比對完整序列或整體結構相似度來尋找潛在蛋白質的方法，我們藉由SRP 來描述蛋白質與小分子的結合環境，並以3D-BLAST 將三級結構片段轉換編碼成一級序列，對目前所有已知蛋白質結晶結構進行快速搜尋，尋找擁有類似SRP 的蛋白質。進一步結合結構比對工具如DALI，標定相似結合環境在潛在蛋白質中的空間位置。 我們蒐集530 個蛋白質－藥物分子結晶結構，共187 種美國食品藥品監督管理局(FDA)核准的藥物，以此建構個別SRP 並對蛋白質結構資料庫(PDB)進行搜尋。對於搜尋相同蛋白質或共結晶相同化合物的結晶結構，SRP 的覆蓋率(recall)分別為80%和54%。針對搜尋整體或區域結構相似的蛋白質，SRP 準確率(precision)可以達到82%，說 明我們的方法在尋找相似結合環境上可以提供可靠的預測結果。同時在本研究中，我們以治療流感的藥物瑞樂沙(Zanamivir)為例，說明SRP 對結合環境的敏感性以及可能如何應用於蛋白質分類問題上，另一實例以治療慢性骨髓細胞白血病藥物Imatinib，說明SRP應用在「舊藥新用」議題上的可能性。 最後，我們建置一網站提供530 個蛋白質－藥物結晶結構所建構的SRP 資訊以及搜尋結果，用以觀察討論187 種藥物可能的潛在結合蛋白質。本研究利用空間相關結構片段(SRP)來尋找相似結合環境的蛋白質，期望能對瞭解蛋白質－化合物交互作用與結合環境的關係、以及探討藥物於舊藥新用或副作用的研究有所幫助。

It is important to understand the potential target proteins for a chemical compound.During the early drug discovery stage, for example, it could avoid the unnecessary developing cost and time by detecting the potential harmful side effects. On the other hand, it could provide the new usages for old drugs. Recently, multiple target drugs give a new paradigm for diseases with complex mechanism such as cancers and diabetes. Therefore, discovering potential target proteins of a given compound is a valuable issue in bioinformatics and drug development. Previous studies indicate that similar compounds enable to bind the proteins with similar binding environment. Researchers usually search similar proteins by aligning the given protein sequence or global protein structure in sequence or structure databases. However, previous works show that in some cases proteins bound the same ligand may not have significant evolutionary relationship in both sequence and global structure but in their binding environments. In this study, we introduce a concept named Space-Related Pharmamotif (SRP) to discover the proteins with similar binding environment in protein databases. SRP is composed of a set of spatially discontinuous peptide segments, which surround the ligand-binding site. Compared with the previous methods of finding proteins with similar sequence or global structure, SRP focuses on protein-ligand interacting environment. By transforming the 3D structure segments into 1D structural alphabet sequences through 3D-BLAST, we can search the potential target proteins with similar binding environment against Protein Data Bank (PDB) rapidly. Furthermore, we use the structure alignment tool, such as DALI, to precisely locate the possible binding environment in these target protein structures. We collect 530 protein-drug co-crystallized complexes, in which contain 187 different FDA-approved drugs. We build SRPs and screen PDB for each protein-drug complex. For searching the proteins with the same UniProt accession number and the same ligand, the recall achieves 80% and 54%, respectively. Proteins classified into the same homologous superfamily of CATH can be predicted with a precision of 82%. Our results demonstrate that SRP provides a reliable performance in searching the potential target proteins with similarbinding environment. We give an example of Zanamivir to describe how SRP can identify slight structural difference of the binding environments between proteins. In another example, we preliminarily discuss the issue of “new use for old drugs” about Imatinib, which is a marking drug known to against disease chronic myelogenous leukemia and gastrointestinal stromal tumor. Finally, we build a web server to represent the SRP information and the searching results from 530 protein-drug complexes for helping to identify the potential binding protein of 187 known drugs. In this study, we supply evidence to present that SRP is reliable for searching the potential target proteins with similar binding environment. In the future, we will develop SRP to be useful to understand protein-ligand interactions and helpful for drug design.