在非完整之蛋白質作用網裡探勘交互作用機制

Abstract

蛋白質在細胞中扮演著不可或缺的角色。透過蛋白質交互作用的研究，可以推敲蛋白質執行的功能，並理解生物體運作的機轉，進而協助藥物的開發。 目前透過電腦計算來探尋蛋白質交互作用的技術，大部份都是試圖從蛋白質的演化、結構、以及物化性質，探勘出與交互作用有關的特性。不過由於這些資訊的殘缺不全，這些方法只能應用於特定的蛋白質家族。 本研究提出一種先進的計算方法來進行蛋白質交互作用的探尋。此方法可分成兩階段：第一階段藉由分析蛋白質交互作用網路，搜索出以相同機制進行交互作用的蛋白質群集。第二階段則利用樣式探勘演算法，在這些群集中挖掘與交互作用機制相關的氨基酸組合樣式，來進行蛋白質交互作用的預測。此方法的優點為只利用蛋白質序列之資訊，可應用於整個蛋白體，並可獲得蛋白質結合之區域。 本方法預測嶄新蛋白質交互作用的精準度(precision)高達79%，並且在蛋白質立體結構上，成功抓出蛋白質實際結合之區域。

Proteins are indispensable to the functioning of cells. Understanding the ways in which proteins interact with one another provides insight to protein function and biological processes. This understanding can help to design drugs to cure malfunctioning proteins. Computational techniques to discover protein-protein interactions typically use classifiers to find commonalities in the evolution, structure, or physiochemistry of the proteins. However, since we lack complete information for these properties for many proteins, these techniques are applicable for a limited collection of protein families. This study uses a two-step methodology of finding groups of proteins that use the same mechanism for interaction by analyzing the network topology then using a pattern mining algorithm to find the amino acids responsible for the mechanism. Since the technique uses only the primary sequence of the proteins, this technique has the benefit of being applicable at a proteome scale and also provides the additional benefit of revealing the site of interaction. These mined patterns are used to predict novel PPI with a precision of 79% percent, and the amino acids that match the patterns are found to be close to the interaction sites in the three dimensional structures of proteins.