肺癌在近年來已經成為全世界主要死因之一,而肺癌中超過百分之八十五都屬於非微小型細胞肺癌。然而,新藥的開發是非常昂貴且費時的過程,因此如何有效找到潛在的藥物治療非微小型細胞肺癌已是一個非常重要的課題。 在我們以前的研究中,我們已經建立了一套機器學習方法,根據蛋白質作用域交互作用、 加權的作用域頻率分數及癌症連結分支度資料以預測癌症蛋白質。在本論文中,我們擴展此項研究,透過AUC參數進一步評估我們所提方法的效能,並且應用到微陣列差異性表達基因尋找潛在的非微小型細胞肺癌基因。我們發展了一套非微小型細胞肺癌的藥物開發流程,整合潛在的非微小型細胞肺癌基因、蛋白質-蛋白質交互作用、生物路徑分析及cMap等資源。最後所推導的藥物並經由實驗驗證其有效性。我們預期這套非微小型細胞肺癌的藥物開發流程也許可以應用到其他癌症藥物的重新定位上。
Lung cancer is the leading cause of death worldwide, and non-small cell lung cancer (NSCLC) accounts for more than 85% of all lung cancer cases. However, the process of new drug development is cost-intensive and time-consuming; therefore, how to effectively search for suitable potential drugs for NSCLC has been a critical issue in biomedical research. In the previous study, we have developed a machine learning method, based on domain-domain interactions, weighted domain frequency score and cancer linker degree data, to predict cancer proteins. In this thesis, we extended the previous study by further evaluating its performance with AUC (area under curve) measure, and applied the machine learning method to predict potential cancer genes from differentially expressed genes from microarray data. We then developed a pipeline to infer potential therapeutic drugs for disease treatment by preforming meta-analysis, and integrated the protein-protein interactions, biological pathway analysis and the cMap resources. Finally, the predicted drugs are investigated by experiments. It is expect that the drug-finding pipeline may be helpful in drug repositioning discovery for other cancer diseases.