液態生物檢體應用在卵巢癌分類與篩檢

An application of serum exosomes as biomarkers in differentiating histological subtypes of ovarian cancer

藍逵原(Lan Kuei-Yuan)

指導教授 ： 張家銘

國立政治大學/理學院/資訊科學系/碩士(2018年)

https://doi.org/10.6814/THE.NCCU.CS.015.2018.B02

摘要

卵巢癌是女性第八常見癌症，並且在婦科癌症中是致死率最高的一種。我的研究期望能找到卵巢癌相關的生物標記，幫助癌症能在早期確診。假設不同的卵巢癌形態會分泌不同的小分子核糖核酸 (miRNAs) 進而影響週遭細胞的表現導致癌化，那藉由比較這些在微環境中的小分子核糖核酸能夠幫助我們判斷病人是否得到卵巢癌。在小分子核糖核酸的研究中，我們使用了45位病人的樣本，其中29位是帶有不同亞型的癌症，另外16位是控制組。在我們所觀察到的2496個小分子核糖核酸中，263個在癌症病人與控制組間表現上有顯著的差異，再藉由機器學習的方式，我們建立了一個可靠的邏輯回歸模型來分辨病人是否得到卵巢癌，以及卵巢癌的那一種亞型。此外，針對具有較強抗藥性的亞型的病人，也對其基因的表現在正常細胞與癌細胞的不同進行研究。研究共有十位癌症病人，以其中六位病人的正常細胞當作控制組，得到有755個基因在兩組之間表現上有顯著的差異。最後，我們發現了許多過去不曾發現的小分子核糖核酸與基因之間的關係，未來可能用做標靶治療的目標。

關鍵字

卵巢癌 ； 小分子核糖核酸 ； 核糖核酸測序 ； 邏輯回歸分析 ； 機器學習

並列摘要

Ovarian cancer is the eighth common cancer in women, and the most deadly gynecologic malignancy. My master project aims to identify candidates of biomarkers, which may be used in early detection of ovarian cancer. We hypothesize that different subtypes of ovarian cancer may secret exosomes carrying different miRNAs play different roles in cell-cell communication in microenvironment. Therefore, we aim to compare the expression profiles of exosomal miRNA in the serum from patients with or without ovarian cancer. Furthermore, we performed RNA-seq for mRNA profiles in the cancer tissue of a most drug-resistant subtype with their paired normal tissue from the same patients. A total of 45 patients were enrolled in this study. Sera from all 45 patients were used in the study of exosomal miRNA, in which 29 samples are cancer patients and the other 16 are non-cancer controls. RNA-seq data was generated from ten patients who had clear-cell ovarian cancer subtypes, six of them have corresponding paired normal tissue. In miRNA, 2496 miRNAs were identified and 263 miRNAs are differentially expressed between normal samples and cancer samples. We construct a reliable machine learning model to classify patient cancer subtypes base on the candidate miRNAs selected by the model. 755 RNAs are differentially expressed between normal samples and cancer samples. Lastly, we found couple unknown predicted miRNA and mRNA interaction, which may further the candidate of targeted therapy in the future.

並列關鍵字

Ovarian cancer ； MiRNA ； RNA-seq ； Logistic regression ； Machine learning

參考文獻

1. Weng,S.-L. et al. (2017) Genome-wide discovery of viral microRNAs based on phylogenetic analysis and structural evolution of various human papillomavirus subtypes. Brief Bioinform.

2. Wang,Z. et al. (2009) RNA-Seq: a revolutionary tool for transcriptomics.Nat Rev Genet, 10, nrg2484.

3. Elias,K. et al. (2017) Diagnostic potential for a serum miRNA neural network for detection of ovarian cancer. Elife, 6, e28932.

4. Yokoi,A. et al. (2014) A combination of circulating miRNAs for the early detection of ovarian cancer. Oncotarget, 5, 89811–89823.

5. Katz,B. et al. (2015) MicroRNAs in Ovarian Cancer. Hum Pathol, 46, 1245–1256.