| 摘要: | 這篇論文的研究目的是提出一些新的電腦圖學技術與計算方法來解決電腦輔助藥物設計的問題。本論文以幾何、能量與活性等三大方向來探討如何使用最小能量原理來改進電腦輔助藥物設計中的分子對接過程之效能，加速藥物設計時程與降低研發成本。從美國疾病管制局2004年的研究報告中可以說明流行疾病盛行，各種藥物設計必需再加快時程與研發數量。在很多的藥物研究報告中也發現，電腦輔助藥物設計最大的挑戰是分子對接過程，在本論文中我們以最小能量原理為解決問題的主軸。首先從幾何搜尋方面著手，本研究是以受體為基礎，除了比較四種不同的受體特性外，並以研究如何快速模擬蛋白質的摺疊開始，我們提出了改良式遺傳演算法來加速接合位置的幾何搜尋；再來以能量為研究重點，在實驗中使用李亞普諾夫函數中的穩定理論來降低接合位置數以便進一步增進分子對接的效能，並且使用NURBS曲線中的插入頂點與權重調整來加速分子系統達到最小能量狀態。最後我們以各種不同的藥物受體模型來做電腦模擬計算，利用最小能量原理判斷出接近全域能量最小的區域之對接狀態的穩定度，並對其各種分子活性進行評估，研究各種分子對接中的各項活性，從中瞭解分子力場各元素的貢獻度，也成功地驗證出本論文所提出的方法。 本研究導入目前實驗準則X-ray及RMSD的標準，也驗證我們的電腦模擬之結果是否在容許誤差範圍內，並提出Michel，David， Denical，Abraham等人的相關研究來佐證我們的研究之基礎。我們在論文中所提到的模擬環境是以AMBER力場及Ullman's演算法為基礎，在模擬的過程中探討每一時刻特徵值λ的物理意義，比較各個不同對接點之表現結果，分析藥物受體模型中的蛋白質的摺疊和各種鍵結力之影響。在經過了蒙地卡羅，退火演算法，遺傳演算法及改良式遺傳演算法等四種最佳化搜尋方法的實驗後可知，我們所提出的改良式遺傳演算法來執行接合位置的搜尋及完成分子對接程序最快，引用Pegg和Camila的兩篇研究報告作比較，得到較接近全體能量最小值及算術過程之收斂時間快1.16小時。本論文亦研究藥物受體互動模型中的親合力係數之測量，除了判斷藥物與受體的關係是增效或對抗，也知悉分子擾動中那一種對接的活性較強。 根據之前我們所出版的研究論文，我們使用創新的WebDeGrator繪圖系統來建立分子對接過程中的電腦模型。使用藥物受體互動模型，以八個配體及HIV 蛋白脢受體為模擬對象做分子對接而得到一些特性結果。最後，在論文中對於最佳解，分子對接及蛋白質摺疊相關議題也將詳細分析與研究。跨越各種不同研究領域，例如結合生物學，資訊科學，系統學及化學等來解決生物資訊的各種新興的議題將是未來最強有力的解決方法。 This investigation presents novel computer graphical and computational schemes for solving the challenges of computer-aided drug design (CADD). The application of the energy minimum to enhance the docking performance of CADD is discussed in terms of three aspects, geometry, energy and activity. American CDC research reports reveal that an increasing incurrence of disease, resulting in a requirement to accelerate drug discovery. However, commercialization of a new drug is extremely complicated. The most significant challenge is the docking procedure in CADD according to previous literature. This study applies the energy minimum theorem to solve the objection. A geometry search is performed and compared with four types in classification of receptors. This work attempts to improve the speed of computer simulations of protein folding of protein, and proposes an improved genetic algorithm to accelerate the binding site search; second, we focused on energy theme. Lyapunov’s stability theorem is adopted to decrease the number of binding sites, thus enhancing the docking performance in computer simulation examples. The knot insertion and modifying weights of NURBS curves are utilized to accelerate the molecular docking system in order to obtain the shortest response route. Finally, various drug-ligand interaction models are employed to compute docking simulation, and energy minimum theorem is used to judge the approach global energy minimum area and docking stability. Various molecular activities are derived at each binding site, and the contribution of every bond and non-bond’s in the force field is observed. As a benchmark is reference for testing docking performances, the error tolerance of computer simulation examples is compared with the X-ray and RMSD experiment standard, and the values obtained by Michel, David, Denical and Abraham’s researches performance. This investigation develops the AMBER force field and Ullman’s algorithm to support the computer simulation environments. The significance of the eigenvalue λ is analyzed at each protein folding, and this study performance has increased by 25 percents compared with various binding sites. Additionally, the protein folding and various bond forces in drug-ligand interaction model are discussed s. Comparing four optimal geometry search methods and referred to Pegg and Camila the two been published paper in benchmark of drug docking database, the improved genetic algorithms are specified to undertake the search binding site and docking, and the global minimum search and the arithmetic convergence time of 1.16hr is achieved. Analytical results indicate that the improved genetic algorithm is better than traditions random methods in terms of processing the geometry graphics operation. Previous published investigations have employed the WebDeGrator system to establish molecular computer modeling for the docking process. This study demonstrates examples in protein folding kinetics and drug docking computations, and successfully applies the Lyapunov function and molecular dynamics to help determine the system stability. Optimal solutions, molecular docking and protein folding kinetics are also discussed herein. This work integrates various research fields to find advanced and novel solutions to problems in bioinformatics. The combination of biology, information, system, and chemistry will be a powerful CADD strategy in the future. |
