區域官能基地圖對藥物最佳化的網路服務：流感病毒神經胺酸酶為例

Abstract

前導藥物最佳化的過程是在藥物發展過程中的一項巨大的挑戰，由於從虛擬藥物篩選或是高通量篩選所尋找到的前導藥物通常都是在微莫耳等級(μM)。而由電腦幫助前導藥物最佳化的方法可以大致上分為三類：藥效基團為基礎、定量結構關係基礎及片段對接方式。然而在前兩類的方式主要被限制於需要一組已知實驗資料的化合物，因而無法尋找到新型態的化合物，而第三類是藉由記分方式去預測結合能量，但通常記分方式是不準確的。因此發展一項新的方法去加速前導藥物最佳化的過程對藥物設計上來說是有很大的價值。 針對上述議題，我們利用實驗室之前所發展SiMMap服務器去進行藥物最佳化的過程。該服務器從目標蛋白質的結合區域位置及已對接化合物(或是共結晶配體)產生區域官能基地圖。區域官能基地圖由許多錨點(Anchor)所組成，每個錨點由三個基本元素所構成：1)由具一致交互作用之胺基酸所構成的結合立體袋槽(binding pocket)；2)複數個虛擬配體構成的官能基組成；3) 結合立體袋槽與官能基之交互作用關係(包含靜電力、氫鍵及凡德瓦力交互作用)。區域官能基地圖描述在結合區域位置內的物化性質與官能基偏好之間的關係。之前的實驗結果也證實當小分子化合物與區域官能基地圖描述的錨點特性高度相符時，通常有高度潛力成為目標蛋白質的抑制劑或促進劑。綜上所述，區域官能基地圖從立體結構，氫鍵、靜電力等官能基來修飾潛力藥物是有幫助的。 在初步成果中，我們應用SiMMap 藉由篩選化合物來推論流感病毒(H1N1)神經胺酸水解酶(influenza neuraminidase)的區域官能基地圖，並採用流感病毒神經胺酸水解酶的三種藥物(瑞樂沙(Zanamivir)、克流感(Oseltamivir)及帕拉米維(Peramivir))來作為驗證。透過錨點內官能基的偏好程度，我們可以模擬上述藥物的發展過程且不需要具有實驗資料的化合物的幫助。再者，錨點能量與從文獻上所收集的神經胺酸水解酶抑制劑的半數最大抑制濃度(IC50)之間的皮爾森相關係數(Pearson correlation)為0.78，代表說化合物有較低的錨點能量可能視更有效的神經胺酸水解酶抑制劑。根據上述結果，我們相信錨點內官能基偏好程度能夠有助於前導藥物最佳化的過程

The optimization process of initial hits is one of the major challenges in drug development because the initial hits found by high-throughput screening or virtual screening are usually at the micromolar level, Computer-aided lead optimization method can be roughly classified into three categories, including pharmacophore-based, quantitative structure-activity relationship (QSAR)-based, and fragment docking methods. However, the major limitations of the two former methods are that they require a set of compounds with experimental data and are not effective to find compounds with new scaffolds. The latter method predicts binding energy by scoring functions, which are often imprecise. Therefore, developing a new method to accelerate the process of lead optimization can provide a great value in drug design. To address the issue, we applied the site-moiety map (SiMMap) server developed by our lab to optimize potency of compounds. The server can derive a SiMMap of a protein binding site from a target protein and its docked (or co-crystallized) compounds. A site-moiety map consists of several anchors, and an anchor includes three basic elements: a binding pocket with conserved interacting residues, the moiety composition of query compounds, and pocket-moiety interaction type (electrostatic, hydrogen-bonding, or van der Waals). The site-moiety map describes the relationship between the moiety preferences and physico-chemical properties of the binding site. Our previous experimental results showed that when a compound highly agrees with anchors of site-moiety map, this compound often activates or inhibits the target protein. As a result, the site-moiety map can be helpful to assemble potential leads by optimal steric, hydrogen-bonding, and electronic moieties. Here, we applied the SiMMap server to infer the site-moiety map of H1N1 neuraminidase by screening compounds, and verified the utility in the optimization process of three neuraminidase drugs (Zanamivir, Oseltamivir and Peramivir). Through the moiety preferences of anchors, we can simulate the development process of these drugs without using compounds with experimental data. In addition, the Pearson correlation coefficient between the moiety energies of anchors and IC50 values of neuraminidase inhibitors collected from literatures is 0.78, suggesting a compound with low anchor moiety energies could be a potent neuraminidase inhibitor. According to our results, we believe that moiety preferences of anchors are useful for the process of the lead optimization.