膀胱癌為泌尿系統中常見的惡性疾病之一，其常見的症狀為血尿，可能可透過肉眼看見或經過尿液檢驗發現是否有顯微血尿。為了解膀胱癌病人其基因之突變與表現量對於病人存活 (及復發) 和癌細胞型態的影響，本研究資料來自美國癌症基因體圖譜計畫 TCGA (The Cancer Genome Atlas)，經過史隆凱特林癌症研究中心 (Memorial-Sloan-Kettering Cancer Center, MSKCC) 計算生物學中心整理的癌症基因資料庫 (Cancer Genomics Data Server, CGDS) 下載，透過高維度迴歸分析之方法，利用如 Lasso、Ridge、Adaptive Lasso 及 Cox Model 處理變數篩選，並比較不同方法之預測結果，以及做風險預測。分析中找到的基因，和文獻中提到和膀胱癌相關的基因做印證，以了解這些基因和膀胱癌的關係。同時可以幫助實驗室後續的尿液試紙檢驗分析，希望可以早一步篩檢出癌症病人，以達到最好的治療成效。另外，本研究也利用 R 程式寫了如何取得和處理基因資料庫資料的函式，以供將來研究者使用。

Bladder cancer is one of the malignant diseases in urinary system. Its common symptoms include hematuria which could be seen through eyes or urine analysis. In order to understand the effect of gene expression and mutation data on subtypes and recurrent event in patients with bladder cancer, we downloaded data from The Cancer Genome Atlas (TCGA) and applied high-dimensional analysis such as LASSO, Ridge, Adaptive Lasso and Cox model to screen gene variables, compare the performance of different models and predict the hazard of each patients. Among the selected gene candidates, we found TP53 and ERBB3 have been published in quite a few papers, which could verify our method. Not only the list of genes could help the lab to perform further analysis but also it could screen out the potential patients in advance. On the other hand, we also wrote some functions to access and deal with gene database in R language, which could be used by other researchers in the future.

論 文 審 定 書 i
論 文 公 開 授 權 書 ii
誌 謝 iii
摘 要 iv
Abstract v
1 研 究 動 機 與 目 的 1
2 資 料 描 述 2
2.1 資料取得 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 資料處理 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.3 變數介紹 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3.1 臨床變數 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3.2 基因資料之變數 . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.4 敘述統計 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3 研 究 方 法 5
3.1 錯誤發現率 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 邏輯斯迴歸(Logistic regression) . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.1 敏感度 (Sensitivity) 與特異度 (Specificity) . . . . . . . . . . . . . 6
3.2.2 ROC 曲線 (Receiver Operating Characteristic curve, ROC curve) 7
3.2.3 約登指數 (Youden’s index) . . . . . . . . . . . . . . . . . . . . . 7
3.3 存活曲線 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.4 Cox 比例風險模型 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.4.1 一致性指標 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.5 正規化之方法 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.5.1 正規化 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.6 五折交叉驗證 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.7 覆蓋率計算 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.7.1 兩個基因之覆蓋率計算 . . . . . . . . . . . . . . . . . . . . . . . . 12
4 研 究 結 果 13
4.1 以癌症亞型為反應變數 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.2 以復發為反應變數 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.3 覆蓋率 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3.1 基因表現量之覆蓋率 . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3.2 基因表現量結合突變基因之覆蓋率 . . . . . . . . . . . . . . . . . . 26
4.3.3 基因表現量結合突變基因移動寬度之覆蓋率 . . . . . . . . . . . . . 29
5 結 論 與 建 議 31
參 考 文 獻 32
A 附 錄 34

[1] Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a
practical and powerful approach to multiple testing. Journal of the royal statistical
society. Series B (Methodological), 289-300.
[2] Cancer Genome Atlas Research Network. (2014). Comprehensive molecular characterization of urothelial bladder carcinoma. Nature, 507(7492), 315.
[3] Cerami, E., Gao, J., Dogrusoz, U., Gross, B. E., Sumer, S. O., Aksoy, B. A., ...& Antipin, Y. (2012). The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data.
[4] Friedman, J., Hastie, T., & Tibshirani, R. (2010). Regularization paths for generalized linear models via coordinate descent. Journal of statistical software, 33(1),1.
[5] Gao, J., Aksoy, B. A., Dogrusoz, U., Dresdner, G., Gross, B., Sumer, S. O., ...& Cerami, E. (2013). Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci. Signal., 6(269), pl1-pl1.
[6] James, G., Witten, D., Hastie, T., & Tibshirani, R. (2013). An introduction to statistical learning (Vol. 112). New York: springer.
[7] Le Cessie, S., & Van Houwelingen, J. C. (1992). Ridge estimators in logistic regression. Applied statistics, 191-201.
[8] Simon, N., Friedman, J., Hastie, T., & Tibshirani, R. (2011). Regularization paths for Cox’s proportional hazards model via coordinate descent. Journal of statistical software, 39(5), 1.
[9] Storey, J. D., & Tibshirani, R. (2003). Statistical significance for genomewide studies. Proceedings of the National Academy of Sciences, 100(16), 9440-9445.
[10] Tibshirani, R. (1996). Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society. Series B (Methodological), 267-288.
[11] Tibshirani, R. (1997). The lasso method for variable selection in the Cox model. Statistics in medicine, 16(4), 385-395.
[12] Zhang, H. H., & Lu, W. (2007). Adaptive Lasso for Cox’s proportional hazards model. Biometrika, 94(3), 691-703.
[13] Zou, H. (2006). The adaptive lasso and its oracle properties. Journal of the American statistical association, 101(476), 1418-1429.