GEMDOCK於虛擬藥物資料庫篩選之功能增進及套膜蛋白與醯亞胺水解酵素之實際應用

Abstract

針對已知結構的蛋白質，應用虛擬藥物篩選的工具在化合物資料庫中尋找潛在的抑制劑，是目前電腦輔助藥物設計廣為使用的方法之一。其目的在於有效地縮短尋找潛在抑制劑的時間，並減低實驗所需的成本。在本篇論文中，除了希望將GEMDOCK全程自動化之外，還希望藉著計分程式的修正以增強GEMDOCK在虛擬藥物篩選上的功能。GEMDOCK的演算法同時具有區域及全域搜尋最佳解的優點，而且修正後的計分程式不但提高了分子鉗合的準確度，並在虛擬資料庫篩選上有明顯的改進，確實減少了偽陽性的數目。本研究先以TK (thymidine kinase)、ER (estrogen receptor)及hDHFR (human dyhydrofolate reductase)為標的蛋白，從化合物資料庫 (ACD、MDDR) 中隨機挑選出990個化合物，再加上已知會與標的蛋白結合的配體各10個，做為測試組，以GEMDOCK預測這1,000個化合物分別與蛋白質鉗合的位置及能量，將結果依能量排序，以觀察GEMDOCK在篩選化合物資料庫上的表現。結果在true hit%達100%時，偽陽性的比例(false positive rate)分別為9.7%(TK)、5.2%(ER antagonists)、21.2%(ER agonists)及8.6%(hDHFR)；若在蛋白質活性區域中重要的胺基酸上，根據10個已知配體的特性來加重計分，則偽陽性的比例分別減少為2.9%(TK)、0.9%(ER antagonists)、1.9%(ER agonists)及2.0%(hDHFR)。在確認了GEMDOCK在篩選上的表現之後，我們將GEMDOCK應用於登革病毒套膜蛋白(Envelope protein)的抑制劑篩選上，登革熱是台灣夏季的流行性疾病，而登革病毒的套膜蛋白則為可能的藥物設計標的。GEMDOCK也被應用在辨識D-hydantoinase的基質或抑制劑上，結果我們發現兩個新的基質，並與生物實驗的結果相符。

Virtual ligand screening is a method broadly used for computer-aided drug design. It will save much time and cost to find potential inhibitors for the target protein with aids of computers. In this thesis, we have developed an automatic tool with a novel scoring function for virtual screening by applying numerous enhancements and modifications to our original techniques, called GEMDOCK. By integrating a number of genetic operators, each having a unique search mechanism, GEMDOCK seamlessly blends the local and global searches so that they work cooperatively. Our scoring function is indeed able to enhance the accuracy during the flexible docking and to improve the screening utility by reducing the number of false positives in the post-docking analysis. First we have verified our program with four screening sets for thymidine kinase (TK) substrates, estrogen receptor (ER) antagonists, estrogen receptor agonists, and human dihydrofolate reductase (hDHFR) ligands. These four screening sets were composed of ten known ligands for each target protein and 990 compounds randomly selected from the ACD or MDDR. The 1,000 compounds of the four screening sets were docked into each target protein and ranked according to their potentials. When the true hit rate was 100%, the false positive rates were 9.7% for TK, 5.2% for ER antagonists, 21.2% for ER agonists, and 8.6% for hDHFR. After adding pharmacological consensuses on important residues and ligand preferences according to the ten known ligands, the false positive rates were decreased to 2.9% for TK, 0.9% for ER antagonists, 1.9% for ER agonists, and 2.0% for hDHFR. After verifying the utility of GEMDOCK on virtual screening, we applied it to identifying potential inhibitors for the envelope protein of dengue virus. Dengue fever is an epidemic disease in Taiwan during the summer. The envelope protein (the PDB entry: 1oke) of dengue virus is a possible target for drug design. Finally, GEMDOCK has been also accessed on identifying new substrate/inhibitors of Agrobacterium radiobacter hydantoinase, which is an industrial enzyme. We have screened candidates for hydantoinase and identify two new substrates evaluated by wet experiments.