整合多重基因體資料以分析腫瘤浸潤免疫細胞與預後之關係

Integrate Multi-omics Data to Explore Association between Tumor-Infiltrating Lymphocytes and Tumor Prognosis

張羿玄(Yi-Hsuan Chang)

指導教授 ： 莊曜宇

共同指導教授 ： 蕭自宏(Tzu-Hung Hsiao)

國立臺灣大學/電機資訊學院/生醫電子與資訊學研究所/碩士(2017年)

https://doi.org/10.6342/NTU201702340

摘要

免疫治療藉著活化病人體內的免疫反應來對抗腫瘤，近年來快速的發展且逐漸被應用在一些癌症之中。部分研究指出腫瘤免疫浸潤以及腫瘤抗原特異性會顯著的影響到免疫治療的效果以及病人的預後，因此為了提升免疫治療的效果，腫瘤免疫浸潤的含量便成為一個評估病人免疫治療效果以及預後的重要指標，如何找出影響免疫浸潤的生物標記也成為現今研究的主要趨勢。
在此論文中我們建立了兩套方法整合常見的體學資訊，包含了基因體、轉錄體、表觀遺傳體以及免疫基因體來觀察多重體學綜合下來對癌症患者中腫瘤免疫浸潤基因群以及預後的影響性。第一種方式是利用基因群分析找出和癌症預後顯著相關的免疫浸潤細胞種類，並且透過聚類分析探討差異表現的基因群可能調控的生物途徑。在第二種方法之中，我們結合腫瘤基因體學以及免疫基因體學，並利用隨機森林分類法篩選出對腫瘤免疫浸潤含量高低具有影響性的特徵，接著透過特徵重要性排序以及功能性註解找出潛在的分子機轉功能。我們將設計出來的兩種方法運用在GEO乳癌患者資料以及TCGA的大腸直腸癌患者資料上，分析腫瘤浸潤免疫細胞與預後之關係。在雌激素受體表達的乳癌患者之中發現有九種免疫浸潤細胞會顯著影響患者預後，並且於其中找出了先天與後天免疫浸潤細胞的互動關係，證實了腫瘤免疫浸潤細胞會共同調控患者之預後。另外在大腸癌患者中，透過特徵篩選以及隨機森林演算法成功鑑別出二十一種免疫細胞浸潤含量的高低，並於兩種免疫浸潤細胞中發現免疫球蛋白受體相關生物功能參與了免疫浸潤細胞含量的調控，顯示了我們結合多種體學特徵能夠更完整的找出影響免疫浸潤細胞含量的生物標記。
總結而言，本實驗系統性的探討了腫瘤免疫細胞浸潤的相關生物標記，並證明免疫細胞浸潤具有預測病人預後的潛力，期盼藉此為免疫治療效益以及病人預後的評估提供幫助，也進一步瞭解腫瘤免疫間的交互作用。

關鍵字

腫瘤免疫浸潤 ； 基因群分析 ； 隨機森林 ； 多重體學 ； 存活分析

並列摘要

Immunotherapy is a treatment that activates immune responses to fight tumors. This therapeutic option has advanced rapidly in recent years and been applied to certain cancers. Previous researches have indicated that both tumor-infiltrating lymphocytes (TILs) and antibody-neoantigen specificity have a great impact on the immunotherapeutic efficacy and the prognosis of cancer patients. Therefore, TILs are confirmed as an important indicator to enhance the therapeutic efficacy. Investigating specific biomarkers for TILs then becomes the current research trend.
We devised two methods to integrate multi-omics data, including genomics, transcriptomics, epigenomics and immunogenomics. Through integration of multi-omics data, we can explore the association between TILs and tumor prognosis. In the first method, we employed gene set approach to identify TILs relevant to tumor prognosis and investigated the biological pathway associated with differentially expressed TILs through clustering analysis. With the combination of the tumor genomics and the immunogenomics, we utilized the random forest classification as the second method to identify the characteristics which affected the quantity of TILs. Thus, we can explore potential molecular mechanisms by ranking feature importance and functional annotation. We applied the methods mentioned above to the breast cancer cohorts from GEO and the CRC cohorts from TCGA and analyzed the association between TILs and the prognosis. The results indicated that 9 TILs subtypes significantly associated with ER+ breast cancer patient prognosis. Furthermore, we revealed the interaction of innate and adaptive immune cells demonstrating the co-regulation of TILs on the breast cancer patient prognosis. On the other hand, we successfully separated CRC patients with high expression TILs from low expression TILs in 21 subtypes by random forest algorithm. We also found the immunoglobulin-related pathways involved in regulation of two TILs subtypes in CRC patients. The results suggest the incorporation of multi-omics data could comprehensively identify the important biomarkers for TILs.
Overall, our study systematically investigated biomarkers associated with TILs and revealed the evidences that TILs were potential predictors for tumor prognosis. The findings are expected to facilitate the evaluation of immunotherapy response and enhance our understanding of tumor-immune interaction.

並列關鍵字

Tumor-infiltrating lymphocytes ； Gene set analysis ； Random forest ； Multi-omics ； Survival analysis

參考文獻

1. Silberberg, Y., et al., Large-scale elucidation of drug response pathways in humans. J Comput Biol, 2012. 19(2): p. 163-74.

2. Halvorsen, A.R., et al., Differential DNA methylation analysis of breast cancer reveals the impact of immune signaling in radiation therapy. Int J Cancer, 2014. 135(9): p. 2085-95.

3. Ritchie, M.D., et al., Methods of integrating data to uncover genotype-phenotype interactions. Nat Rev Genet, 2015. 16(2): p. 85-97.

4. Ansell, S.M., et al., PD-1 blockade with nivolumab in relapsed or refractory Hodgkin's lymphoma. N Engl J Med, 2014. 372(4): p. 311-319.

5. Nanda, R., et al., Pembrolizumab in patients with advanced triple-negative breast cancer: phase Ib KEYNOTE-012 study. J Clin Oncol, 2016. 34(21): p. 2460-2467.