以官能基為基礎之區域官能基地圖

Abstract

了解生物體內生化反應中蛋白質與配體間交互作用的機制，將對藥物設計相關研究有所助益。目前虛擬藥物篩選是電腦輔助藥物設計中相當有效率的策略，能與實驗互補進而有效降低藥物開發成本。由於蛋白質結構的數量快速增加，虛擬藥物篩選也變得更為重要，然而由於蛋白質與配體之間交互作用尚未被完整地了解，導致在預測潛在藥物時常常不準確。 現今大多數的分子嵌合程式是以能量為基礎計分，而這樣的方法經常忽略蛋白質與配體結合區域中的關鍵交互作用區。為了解決此問題，實驗室於先前發展了以化合物為基礎之區域官能基地圖，區域官能基地圖利用蛋白質結晶結構以及大量(大於1000個)共結晶或與其嵌合的化合物來描述蛋白質結合區域中官能基偏好以及物理化學特性兩者間的關係，此區域官能基地圖可用來了解蛋白質與配體在結合區域中產生的關鍵交互作用以及其作用機制。然而在建立區域官能基地圖前需要將目標蛋白質與上千個化合物執行分子嵌合步驟，而這個過程需要花費大量時間。此外，以化合物為基礎之區域官能基地圖的結果容易受到所嵌合化合物之官能基組成影響，導致在統計官能基偏好時會侷限於特定官能基。 針對這兩項議題，我們提出一個新的概念：建立以官能基為基礎之區域官能基地圖，經由統計1382個FDA認證藥物與6163個生物體代謝物，發現34個經常出現在藥物與生物代謝物中的官能基，再利用這些官能基直接與目標蛋白質執行分子嵌合步驟，而無需嵌合大量化合物，可大幅減少執行時間，而且由於所使用的官能基資料集是經由大量統計現有藥物以及生物代謝物而來，這些官能基可能會是藥物作用或是影響生物體內生化反應的關鍵因素。 我們初步將以官能基為基礎之區域官能基地圖測試在5個重要疾病之目標蛋白質上。我們的實驗結果顯示透過嵌合官能基所識別出的錨點常常位於蛋白質與配體結合區域中的關鍵交互作用區，統計結果指出有82%的錨點與這5個目標蛋白質的受質結合有關，而且98%錨點中的胺基酸在演化上具有高度的保存性，推測這些錨點在生物體內生化反應中扮演重要的角色。此外，符合越多錨點數目的化合物通常具有較佳之作用活性。我們相信以官能基為基礎之區域官能基地圖將能幫助藥物發展、藥物最佳化以及快速地了解蛋白質與配體間交互作用之機制。

Understanding the mechanism of protein-ligand interaction is helpful for drug design. Currently, the virtual screening technique is widely used to predict protein-ligand interactions for reducing the cost and time of drug development. In addition, the rapid increase in the number of protein structures has made the success of virtual screening. However, the accuracy of virtual screening remained intensive. One of the major reasons is our incomplete understanding of protein-ligand interactions involved in biological functions and the imprecise scoring functions. Scoring functions used in virtual screening often ignore key interactions between moieties of compounds and pockets of protein-binding sites, leading a low hit rate. To address this issue, our lab developed a new method, namely compound-based site-moiety map (compound-based SiMMap), to understand the mechanism of protein-ligand interactions and identify the key interactions. A SiMMap utilizes protein structures and numerous docked compounds to describe the relationship between the moiety preferences and the physico-chemical properties of binding site. A SiMMap is composed of several anchors, and each anchor includes three elements: (1) binding pockets (a part of the binding site), (2) moiety preference of the pockets, and (3) pocket-moiety interaction types. SiMMap provides clues to understand key interactions in protein-ligand binding site and their mechanism. However, constructing a SiMMap requires at least 1,000 docked compounds, which is a time-consuming procedure. In addition, the compound-based SiMMap may be biased by moiety compositions of docked compounds. To address these two issues, we propose a novel method namely moiety-based site-moiety map. We firstly identified the 34 most common moieties in 1,382 FDA-approved drugs or 6,163 metabolites. Then, we replaced the compound docking procedure by docking the 34 relevant moieties to save time. Furthermore, the anchors of the moiety-based site-moiety map could be useful to drug discovery and lead optimization because the moieties are the key features of drug actions and metabolisms. We initially tested moiety-based site-moiety map on five important disease target proteins: (1) Thymidine kinase, (2) Estrogen receptor, (3) Shikimate kinase, (4) iii Dihydrofolate reductase and (5) Rho-associated protein kinase 1. We then examined the anchors of the moiety-based site-moiety maps derived by the docked moieties by biological functions or binding mechanisms. Our results reveal that the anchors often located in key interaction areas of protein-ligand binding sites. For example, in the five target proteins, 82% of anchors are involved in the substrate binding or inhibitor binding, and 98% of anchor residues are highly conserved. These suggest that the anchors may play important roles in biological functions and drug design. In addition, we found that the compounds matching more anchors often have better activities. We believe the moiety-based site-moiety map is useful for drug development, drug optimization, and understanding the mechanism of protein-ligand interaction.