在這篇論文中,我們的目的是針對選定的目標蛋白建立可辨識高活性化合物的藥效基團模型來進行電腦輔助藥物設計。選定的兩個蛋白在醫學上表徵為下:人類二氫乳清酸脫氫酵素是一個與癌症和自體免疫及發炎等疾病有高度相關的蛋白酶。血栓素受體蛋白藉著激活血栓素來促進血小板凝聚反應,但過量活化則會導致血栓塞和心血管疾病。基於這兩個與人類多種疾病相關蛋白的重要性,我們利用一群已知生物活性的化合物來建構藥效基團,並使用成本函數分析、費雪隨機分配檢定及命中率預估測試來驗證所建立模型的預估品質及信心強度。數據結果顯示我們建立的藥效基團模型具有優良的預測能力。由於人類二氫乳清酸脫氫酵素的結晶結構已從實驗上被取得,然而血栓素受體蛋白的結晶結構尚未被解出,我們便針對這兩個蛋白建立不同的工作流程。使用所建立出的模型,結合里賓斯基五規則及受體配體對接模擬進行化合物資料庫的虛擬篩選,找出可能成為候選藥物的化合物。經過篩選後,我們找出了一百五十五個針對人類二氫乳清酸脫氫酵素的候選抑制劑,以及五千六百二十一個針對血栓素受體蛋白的候選抑制劑,這些抑制劑未來可提供給其他研究團隊進行後續的藥物實驗。
In this research, our objective is to build the pharmacophore models for selected target proteins that can identify inhibitors with high biological activities and to execute computer-aided drug design. Human dihydroorotate dehydrogenase (hDHODH) is an enzyme which is strongly correlated with certain cancers and autoimmune and inflammatory diseases. Thromboxane A2 receptor (TP) promotes platelet aggregation when activated by thromboxane A2, but over activation of TP may lead to thrombosis and other cardiovascular diseases. Due to the importance of these two proteins related to many human diseases, we use them as targets to build the hypotheses based on a set of known inhibitors, and then use cost function analysis, Fischer’s randomization and goodness of hit test to validate the quality and the confidence of statistical significance of our models. The results show that our models have excellent prediction ability. According to the crystal structures have been solved or not, we can construct different workflows for hDHODH and TP. Consequently, the pharmacophore model, Lipinski’s Rule-of-Five and CDOCKER docking program were integrated into a workflow for the discovery of potential inhibitor candidates from database. Through these workflows, 155 candidates for hDHODH and 5,621 candidates for TP are retrieved for further study.