Biomarker Discovery for Hepatitis C Progression using Machine Learning Techniques

简介（由申请人提供）：丙型肝炎病毒（HCV）感染在普通人群中的发病率正在显著增长，这是一个值得关注的问题。病毒感染的进展以及治疗方式严重依赖于患者的纤维化阶段。因此，如果我们要开出适当的治疗方案，我们需要能够清楚地区分与HCV感染相关的肝纤维化的五个阶段。对于HCV感染如何在晚期纤维化患者中发展为肝癌尚不清楚，因此本项目的一个重要目标是发现早期肝癌检测的生物标志物。将比较与HCV感染相关的纤维化的不同阶段，以发现哪些蛋白质和代谢物有差异表达；其目标将是开发用于纤维化阶段测定的生物标志物面板。此外，将检查患者对治疗的蛋白质组学和代谢组学反应，以确定哪些个体对治疗反应良好。肝癌患者还将与HCV感染患者进行比较，以开发用于检测早期肝癌的生物标志物面板。为了开发这些生物标志物，将对感染患者的血清和尿液样本进行表面增强激光解吸/分离和代谢组学实验。我将使用经典的数据分析方法，如主成分分析、层次聚类、神经网络和逻辑回归来创建一个一级生物标志物面板。在第二阶段，我将使用更复杂的机器学习技术，如核方法和支持向量机（SVM）。我的研究将特别侧重于在支持向量机方面取得进展，以创建用于HCV感染和疾病进展的高质量生物标志物面板。具体目标是：1)使用经典技术分析HCV数据，以识别在纤维化各个阶段差异表达的蛋白质和代谢物；2)利用支持向量机对数据进行分析，更准确地划分纤维化分期；3)改进SVM方法，获得更可靠的肝纤维化分期诊断；4)使用改进的支持向量机技术分析患者对治疗的反应，以确定哪些患者更有可能对治疗产生反应；5)从支持向量机的应用中开发区分癌症和非癌症HCV患者的标记物。

DESCRIPTION (provided by applicant): The incidence of Hepatitis C Virus (HCV) infection in the general population is growing significantly which is of great concern. The progression of the viral infection as well as treatment modality critically depends on the patient's stage of fibrosis. Thus, we need to be able to clearly distinguish between the five stages of liver fibrosis associated with HCV infection if we are to prescribe the proper treatment. Little is known about how HCV infection progresses to liver cancer in patients with advanced fibrosis, so one important goal of this project is to discover biomarkers for the detection of early stage liver cancer. The different stages of fibrosis associated with HCV infection will be compared to discover which proteins and metabolites are differentially expressed; the goal of which will be the development of a biomarker panel for fibrosis stage determination. Also, patient's proteomic and metabonomic responses to therapy will be examined to determine a priori which individuals will respond well to therapy. Patients with liver cancer will also be compared to HCV infected patients to develop a biomarker panel for the detection of early stage liver cancer. For the development of such biomarkers, Surface Enhanced Laser Desorption/ lonization and metabonomics experiments will be conducted on serum and urine samples from infected patients. I will use classical methods of data analysis such as principal components analysis, hierarchical clustering, neural networks, and logistic regression to create a first order biomarker panel. In phase two, I will utilize more sophisticated machine learning techniques such as kernel methods and support vector machines (SVM). My research will especially focus on making advances in SVMs to create high quality biomarker panels for HCV infection and disease progression. The specific aims are to 1) analyze the HCV data using classical techniques to identify proteins and metabolites which are differentially expressed in the various stages of fibrosis; 2) analyze the data using SVMs to more accurately classify the stage of fibrosis; 3) improve the SVM methodology to obtain a more reliable diagnostic of liver fibrosis stage; 4) analyze patient response to treatment using improved SVM techniques to determine which patients are more likely to respond to therapy; and 5) develop markers that distinguish cancer versus non-cancer patients with HCV from applications of SVMs.