基因體醫學及生技研發之生物資訊核心設施-(子計畫四)結構生物資訊(II)

Abstract

在此計畫【基因體醫學及生技研發之生物資訊核心設施：結構生物資訊】，我們將對提供結構分析工具、資料庫之網路服務、教育訓練及研討會，促進基因體醫學及生物科技之研究。我們所提出的核心設施建設在目前結構生物資訊核心基礎上，是現在台灣唯一的記算結構生物核心設施。我們目前我們新開發的服務包括： 1. 蛋白質結構預測 2. 蛋白質結構快速比對 3. 結構分析工具 如：「環結拓樸」特質蛋白質分析工具 4. 蛋白質與蛋白質作用結構資料庫 而我們也將繼續維持目前的資料庫的服務，包括雙硫鍵蛋白質資料庫、結構亂度資料庫。我們也繼續維持目前的鏡像站的服務包括PDB, SCOP and PredictProtein，這些是結構生物一些最重要的網站。在台灣，本核心是唯一提供如此結構資料庫的服務。在分析工具方面，我們將提供下列服務：自動化蛋白質結構預測伺服器、蛋白質－蛋白質／配位子接對伺服器與一般結構分析工具。總括而言，我們的工作包括：一般服務、技術發展與合作服務。我們將貢獻65%在一般服務，連續三年，在合作服務，我們在第一年將貢獻10%的工作力，第二年將貢獻15%，第三年將貢獻20%；在技術發展上，我們在第一年將貢獻20%的工作力，第二年將貢獻15%，第三年將貢獻10%。 PS2是我方開發的自動化蛋白質結構預測工具目前針對蛋白質結構預測，我們實驗室開發了一個蛋白質三級結構預測的工具(PS)2，它使用一致性的策略在蛋白質模版的選取及比對上都有不錯的結果。然而目前蛋白質二級結構預測的成功率以高達80%，因此我們實驗室希望能開發一個新的蛋白質比對工具，它同時能考慮氨基酸及二級結構的資訊，藉此能偵測到蛋白質較遙遠的同源關係。我們預期這將會提升我們在蛋白質三級結構預測的成功率。 另外我們也成功開發出新的工具研究蛋白質分子結構與擾動的相互關係，蛋白質固定點模型。這個模型在實行上提供了一個較方便的方法來計算蛋白質動力學特性，因為它可以直接從計算蛋白質空間幾何的型態得來而不需要複雜的軌道積分或是艱難的矩陣運算，所以此方法比分子動力學模擬或是正規模式分析來的有效率。相較於傳統的分子動力學模擬方法，此模型可節省近10000倍的時間複雜度以及空間複雜度。

In this project Bioinformatics Core for Genomic Medicine and Biotechnology Development –Structural Bioinformatics, we will offer bioinformatics service for both academic and industrial researchers in providing structural-based analysis tools and databases web service, educational work shop program and research conferences, which relevant to the users' research topics in genomic medicine and biotechnology. The current service facility is based on the foundation the current structural bioinformatics core, which is the only bioinformatics core completely dedicated to computational structural biology in Taiwan. We will provide structural analysis tools and database such as follow： 1.protein structure prediction. 2.fast structure alignment. 3.structure analysis (for example, detection of knots in proteins). 4. protein-protein/ligand docking tool GemDOCK and other prediction tools. Furthermore, we will also provide analysis tools and databases in proteomic analysis or genomic analysis, such as protein subcellular localization predictor, phosphorylaton site prediction, database of mircroRNA etc. We will also upgrade the current databases such as the disulfide-pattern protein database and structural entropy database, and continue providing mirror service of some of the most important structural databases such as PDB, SCOP and PredictProtein to the local users. In general, our efforts will be dedicated to three categories: the routine service, technology development and the collaborative research service. We will provide two types of service to users: the first type includes the routine web-based service that will automatically return results for queries submitted by users; the other type of services require special customized assistance from us in carrying out computational intensive operations. For example, our structure prediction server will automatically return 3D structures for a limited number of query sequences submitted by users; however, when the query sequences are, say, all sequences of certain pathogenic genomes or genes related to certain cancer that need comprehensive computation and customized/novel tools, we will provide customized assistance for the users. In general, we will dedicate 65% of our effort to the routine service for each year; as for the technology development, we will dedicate 20% of our effort for the 1st year, 15% 2nd year and 10% the third year, and the rest effort will be dedicated to the collaborative research PS2 is the automatic protein structure prediction server. It uses an effective consensus strategy both in template selection, and target–template alignment. Recently, predicting the secondary structure of a protein has been more successful. Our aim is to propose a new alignment algorithm that both consider the amino acid information the secondary structure information. We expect it will improve the accuracy of our protein structure prediction. We also develop a new method ro research the correlation of fluctuations of proteins. This method, referred to as the protein fixed-point (PFP) model. In practice, this model provides a convenient way to compute the average dynamical properties of proteins directly from the geometrical shapes of proteins without the need of any mechanical models, and hence no trajectory integration or sophisticated matrix operations are needed. As a result, it is more efficient than molecular dynamics simulation or normal mode analysis. Comparing PFP model with traditional molecular dynamics simulation method. PFP model saves 10000 folds time complexity and space complexity.