生物系統內分子交互作用及生化路徑之大規模分析---(總計畫與子計畫一):以生化網路演化關係研究分子交互作用與生化路徑(I)

Abstract

第一期生物資訊跨領域研究計畫(2005-2008)中，本團隊在過去二年半已獲得可觀的學術(40 篇期 刊論文)及實際應用成果(附錄A)。本跨領域整合計畫將延續上期研究成果，首要目的在利用已建立生 物分子序列－結構－功能－生化路徑間的關係，深入研究分子間交互作用機制、演化關係並建立基因 調控網路模型。基因序列、蛋白質及其生物功能間的複雜關係與疾病、藥物、生醫研究密不可分，生 化反應網路由生物分子的交互作用組成，這是蛋白質體學中亟待解決的問題。本計畫的研究成果，有 助於高精確度地分析、預測蛋白質交互作用及生化網路的構成方式，以及未來藥物的發展。簡言之， 即是以序列、結構資訊與演算法為基礎，解析生物系統(biological systems)，並以實驗驗證之。 本計畫包括四個子計畫，分別是以生化網路演化關係研究分子交互作用與生化路徑(子計畫一)、 智慧型最佳化方法用於基因網路的重建與分析(子計畫二)、建立大型分子交互作用資料庫系統與資料 探索(子計畫三)，以及蛋白質於細胞位置之預測與蛋白質間的交互作用：由基因體序列到蛋白質功能(子 計畫四)。四者的緊密結合，可以涵蓋由基因、蛋白質、訊息傳遞路徑(signaling pathway)、代謝路徑 (metabolic pathway)到基因調控網路(gene regulatory networks)間各層次的完整研究。 本跨領域計畫研究目標條列如下： 一. 找出會發生交互作用的功能區塊、其交互作用模式，以及跨物種間共有的蛋白質-蛋白質與蛋白質 -DNA、蛋白質-RNA 交互作用基本模組(consensus patterns)。並發展以演化關係與跨物種比對為基 礎的測量函式(evolutionary and multiple species-based scoring functions)預測蛋白質間及蛋白質-核酸 分子間的親合力(binding affinity) (子計畫一、二、四)。 二. 建置超大型生物分子交互作用(molecular interactions)及生化路徑(pathways)資料庫(database)，涵蓋 基因、蛋白質、生物分子交互作用、生化路徑、基因調控網路，進行收集及整合，並自動更新。 這個資料庫系統將具備完整性、支援物種完整以及不重複等特性(子計畫三、二、一及四)。 三. 以多角度發展合理的分子嵌合(molecular docking)預測方法，這套方法可用於預測蛋白質間交互作 用，運用在醫療上可用在抗原-抗體結合之抗原決定部位(epitope)，也可以預測immunogenicity of MHC class I and II binding peptides，並以此為基礎研究免疫系統(子計畫二、一)。 四. 以演化式最佳化方法等機器學習(machine learning)方法重建基因調控網路，同時針對實驗資料不足 的情況，提出有效之改善方法。並利用子計畫一及二提出的分子交互作用模型與子計畫三資料庫 得到之知識，使重建的網路模型符合真實的生物系統。 五. 以已發展完成的在蛋白質細胞內位置預測工具掃描細菌與病毒蛋白質體，由病原體基因可能在宿 主表現的細胞位置進行相關蛋白質間的交互作用分析，藉此更深入探討病原體基因作用及表現的 機制，並以實驗驗證之(子計畫四、一)。 六. 針對疾病(disease)、癌症治療(cancer therapy)、環境微生物及生質能源等相關議題中的特定生物系 統機制，以多角度探討蛋白質-蛋白質、蛋白質-DNA 及蛋白質-RNA 之間的交互作用，以及這些交 互作用在生化路徑及調控網路中扮演的角色，並以實驗驗證之(子計畫三、二、一及四)。

During the first-term “Program for Interdisciplinary Research Projectin Bioinformatics (2005-2008)”, we have achieved specific goals and published ~40 journal papers (see appendix A). Based on these achievements (correlations between sequence, structure, function, and biochemical pathways), the primary theme of our integrated project is to investigate molecular interactions and evolution relationships, and build various pathways (i.e. signaling pathways and metabolic pathways) and networks (e.g. gene regulatory networks). Disease, drug and biomedical researches correlated closely with the complications among gene sequence, protein and biological functions. The fruitful outcomes of our project will be in accurately analyzing and predicting molecular interactions and biochemical network components, moreover, drug developments. Furthermore, our predictions will be immediately confirmed by cellular and viral experiments. This project consists of four subprojects: network evolution for studying molecular interactions and pathways (subproject 1); intelligent optimization methods for reconstruction and analysis of gene networks (subproject 2); data management and exploration of molecular interactions and pathways (subproject 3), and protein subcelluar localization prediction and protein-protein interactions: from genomic sequences to protein functions (subproject 4). This integrated project studies molecular interactions and networks on genes, proteins, signaling pathways, metabolic pathways, and gene regulatory networks. The specific aims of our project are listed as follows: 1. To identify the interactions and binding models of protein functional domains, as well as the protein-protein, protein-DNA and protein-RNA consensus patterns across multiple species. We will develop evolutionary and multiple species-based scoring functions to predict molecular binding affinity and interacting models by using evolutionary relationships across species. Based on these predicting models and scoring functions, we can infer potential interactions from known interaction networks. In addition, we are able to compare interaction networks across different species to discover the conserved interaction pathways (Subprojects 1, 2 and 4). 2. To establish an integrated database system, including sequences, structures, structural visualized interactions, molecular interactions, biochemical pathways, and gene regulatory networks. The database, which will be updated and integrated automatically, is designed as a comprehensive, broad species coverage and non-redundant molecular interaction repository (Subprojects 1, 2, 3 and 4). 3. We will develop methods for predicting molecular interacting sites (e.g. protein-ligand, protein-protein, and protein-DNA) and molecular recognitions in integrated ways (Subprojects 1, 2 and 4). These methods will be used to practical applications, such as immunogenicity of MHC class I and II binding peptides, structure-based drug screening (e.g. dengue virus and cancer targets) and protein-protein interactions (Subprojects 1 and 2). 4. To reconstruct the gene regulatory networks and to revise deficiency of experimental data (noise and insufficient data) by optimization methods, such as genetic algorithms, support vector machines, and neural networks. We will fit the reconstructed gene regulatory networks to the natural biological systems based on the molecular interactions models (Subprojects 1, 2 and 3) and the biological databases. 5. To identify protein locations and interactions of bacterial and viral proteomes by the well-developed tools, (e.g. CELLO and 3D-partner). Based on the localizations and interactions of host and viral proteins within the host cells, we will be able to predict and investigate the extensive interactions involved in disease mechanisms. These predictions can be verified by experiments (Subprojects 1 and 4). 6. To cooperate with biological experiments to evaluate our computational results on some important biochemical interactions and networks, such as the mechanisms in diseases, cancer therapy and production of biomass energy. We will use multiple strategies to explore protein-protein, protein-DNA and protein-RNA interactions, and find out the roles of these interactions in these networks (Subprojects 1, 2, 3 and 4).