以演化核心模組與蛋白質交互作用網路推導全面性模組

Abstract

在相同時間和空間尺度下，在細胞中分子間精確地聚集且協同作用對於生物程序是不可或缺的。模組（module）是由一群具有高度連結並執行特定生物功能的蛋白質所組成。透過分析模組家族（module families）的組成特性，發現其模組中的核心組成（core component）在調控生物功能時扮演著重要的角色，核心組成乃指跨多個物種的同源蛋白質交互作用家族。因此，我們提出一個新概念：「全面性模組（comprehensive module）」，透過量化模組在不同物種中的演化及蛋白質交互作用網路中不同的時間/空間下的動態行為以探討模組的動態。 在這篇論文中，我們提出一個策略，透過同時考慮多個物種中的演化核心組成以及單一物種中動態的蛋白質交互作用網路（protein-protein interaction network）來推導全面性模組。一個全面性模組包含核心組成（core components；在演化及不同的時間/空間中都具高度保留性的蛋白質/蛋白質交互作用）與周圍組成（ring components；隨演化及不同的時間/空間來變動的蛋白質/蛋白質交互作用）。首先，我們從CORUM資料庫收集現有實驗紀錄之模組作為起始點來建立其模組家族。透過我們先前提出的蛋白質交互作用家族（PPI family），可用來建立人類的蛋白質交互作用網路及計算此模組家族中蛋白質與蛋白質交互作用的演化的程度。此外，我們發展一種貪婪法來找尋全面性模組，主要是透過加入周圍的蛋白質或移除內部的蛋白質來檢測連通-演化內聚性（connectivity cohesiveness）與演化程度（evolutionary scores）的變化。連通-演化內聚性是用來評估蛋白質聚集的連通密集度，以及蛋白質與蛋白質交互作用在NCBI生物分類資料庫中6個生物分類的演化程度。最後，最高分的候選模組即為全面性模組。 進一步驗證424個全面性模組的特性，結果顯示其蛋白質間往往具有高度連結性，且在Gene Ontology的註解上傾向執行相似的生物功能。此外，我們發現全面性模組的核心組成往往是由生物體生存所需的必需蛋白質，同時也經常是蛋白質交互作用網路中的樞紐（hubs），樞紐經常扮演與疾病相關的重要角色。透過計算超過7000個微陣列資料集（microarray datasets），實驗結果指出模組中的核心組成相較於周圍組成在統計上有顯著地共表現程度（co-expression）。意謂當一個模組進行生物功能時，核心組成在不同的時間/空間中確實傾向共同表現，而周圍組成則在不同的時間/空間中呈現動態變化。這些結果顯示我們找到的全面性模組有助於同時在演化與不同的時間/空間中模擬模組的動態行為。我們相信此演算方法及全面性模組的概念將有助於用來了解細胞行為與疾病機制。

Dynamically assembling and cooperating of proteins in time and space is essential to execute biological processes in a cell. A module is a group of proteins that are often highly connected and perform a certain kind of biological functions. The core components, defined as the conserved PPIs and proteins, of a module family often play an important role to regulate biological functions. Here, we proposed a new concept “comprehensive module” to explore the module dynamic by quantifying both module evolution across multiple species and dynamic behavior in time and space from protein-protein interaction (PPI) networks. In this thesis, we propose a strategy to infer the comprehensive modules by considering both evolutionary core components across multiple species and dynamic PPI networks in one species. A comprehensive module consists of core components and ring components representing the conserved and dynamic PPIs/proteins, respectively, in evolution and different time/space. First, we used an experimental module recorded in CORUM database as the seed to infer its module family. Based on our previous PPI families, we built the human PPI network and computed evolution scores of the PPIs and proteins in a module family. Furthermore, we developed a greedy approach to identify the comprehensive module candidates by adding or removing a protein to measure their connectivity cohesiveness and evolutionary scores. The connectivity cohesiveness measures the protein connectivity density and the evolutionary score quantified the PPI and protein conservation in six taxonomies based on NCBI taxonomy database. Finally, we identified the module candidate with the highest score as a comprehensive module. We examined the properties of 424 identified comprehensive modules that are often high connectivity and their protein members often perform similar biological functions based on Gene Ontology analysis. In addition, we found that the core components of comprehensive module are often the essential proteins (genes) of an organism and are the hubs, which often play the key role for diseases, in the PPI network. By computing gene expressions over 7,000 microarray datasets, experimental results show that the core components are statistically significantly co-expression than the ones of the ring components. It means that the core and ring components are actually consensus and diverse in different time/space, respectively, when a module performing biological functions. These results show that our derived comprehensive modules should be useful to model module dynamics both in evolution and different time/space. We believe that our computational strategy and comprehensive modules can be used to understand cell behaviors and disease mechanisms.