以環境─官能基關係開發酪胺酸激酶抑制劑

Abstract

酪胺酸激酶(Tyrosine kinases)是一個在518種蛋白激酶中具有90個成員的蛋白激酶家族，在訊息傳遞上扮演著重要角色和當作許多疾病的重要治療靶點。其中已通過美國食品藥物管理局的26個蛋白激酶藥物中，有19個是屬於酪胺酸激酶抑制劑，且用來治療許多癌症。在發展蛋白激酶的選擇性抑制劑是一個緊急且重大的挑戰，因為在酪胺酸激酶成員彼此間具有一個演化上高度保留的三磷酸腺苷結合區域。先前我們的研究團隊已提出一個KIDFamMap資料庫去了解1,208個激酶─抑制劑家族(KIFs)和保留性的KIDFamMap錨點，代表著在激酶上結構子區與激酶抑制劑上的官能基間有保留的作用關係位置，這些關係可以說明抑制劑的選擇性和結合機制。目前已有許多大規模激酶與抑制劑活性資料和多達37,000種抑制劑陸續被發展和發表出來，這些資料提供一個新的機會就是透過激酶上結構子區和化合物上官能基的關係可以去了解蛋白激酶抑制劑選擇性和結合機制。 本論文中，我們提出了環境─官能基關係(EMRs)為了去了解在激酶上的結合子區與化合物官能基間的結合喜好性。我們發展一個EMRscore用來定量環境和官能基在蛋白激酶子區和抑制劑結構上的結合喜好程度。根據目前已超過的200,000蛋白激酶與抑制劑關係對與1,208家族抑制劑家族內，我們找出445 個環境和51個官能基所組成的 18,361對環境─官能基關係。我們的實驗結果表明EMRs關係不僅可用於描述激酶家族間環境與官能基的最佳結合特性。也可以用來探索激酶抑制劑的選擇性。例如：EMRs在大部分激酶上的特定結構子區(例如Adenine區域)，這個區域靠近ATP上之腺苷官能基結合區域，與抑制劑結構上的環形的官能基和具有可以與激酶樞鍊作用區形成氫鍵作用力的氫鍵結合特性之原子(像是氮原子和氧原子)間具有高表現量。此外，我們的EMRs也可以用來說明在蛋白激酶間保留區域上的胺基酸形成點突變所造成的與抑制劑的活性改變。另外為了評估EMR在發現酪胺酸激酶抑制劑的能力，我們使用SVM(Support Vector Machine)去預測大規模實驗數據的抑制活性(含有300個蛋白激酶與178抑制劑)。根據SVM預測模型，在交叉驗證數值和獨立測試數值分別為94.38%和73.09%。基於這些結果，顯示說 EMRs在反應蛋白激酶與抑制劑的結合喜好、蛋白激酶抑制劑的活性和發展新的酪胺酸激酶抑制劑是非常有用的。

Tyrosine kinases, which are a protein family contains 90 members among 518 protein kinases, play important roles in signaling pathways and are promising therapeutic targets for many diseases. Among 26 US FDA-approved kinase drugs, 19 are tyrosine kinases inhibitors. Discovering selective kinase inhibitors is an emergent and challenging task because kinases share an evolutionary conserved ATP-binding site. We have proposed KIDFamMap to explore 1,208 kinase-inhibitor families and the consensus KIDFamMap anchors, which represent conserved interactions between the kinase subsites and consensus moieties of inhibitors, for kinase inhibitor selectivity and mechanisms. As the increasing number of large-scale kinase profiling data and over 37,000 kinase inhibitors, it provides a new opportunity for studying kinase inhibitor selectivity and binding mechanisms through the relationships between kinase subpockets and their moieties. In this thesis, we propose the environment-moiety relationship (EMR) to explore the binding preferences between kinase subpockets and the moieties of kinase inhibitors. We developed an EMRscore to quantify the binding preference between kinase subpockets and the moieties of kinase inhibitors. According to over 200,000 kinase-inhibitor relationships and 1,208 kinase-families, we inferred 18,361 EMRs from 445 environments and 51 moieties. Our experimental results show that EMRs not only explore the binding preferences of kinases and inhibitors, but also reflect the activity of kinase inhibitors. For example, the EMRs of the subsites (the ATP1 anchor) of most kinases, near the adenine of ATP, highly prefer to the ring moiety with hydrogen bonding atoms (e.g. nitrogen and oxygen) to form hydrogen bonds with kinase hinge regions. In addition, our EMRs are able to reflect the activity fold changes of point mutations on the conserved regions. Furthermore, to evaluate the ability of EMRs for discovering the inhibitors of tyrosine kinases, we utilized support vector machine to predict the inhibitions on the large-scale profiling (300 kinases and 178 inhibitors). The accuracies of our SVM models for cross-validation and independent set are 94.38% and 73.09%, respectively. These results show that the EMRs are useful to elucidate the binding preferences, the activity of kinase inhibitors, and discover new inhibitors for tyrosine kinases.