Statistical Methods for Predicting Survival Outcomes from Genomic Data

简介（由申请人提供）：Annette Molinaro博士是耶鲁大学医学院流行病学和公共卫生系生物统计学助理教授。在来到耶鲁大学之前，Molinaro博士是国家癌症研究所的癌症预防研究员。她的长期职业目标是发展统计和计算方法，阐明癌症发病机制，用于癌症的预防、诊断和治疗。为了实现这一目标，她概述了两个需要加强的领域：1)她在功能基因组学方面的知识，特别是与致癌有关的知识；2)精通计算机编程，能够从大量的数据结构中搜索和提取相关信息。全面了解致癌背后的生物学机制，以及实施新颖统计方法所需的先进计算技能，将推动Molinaro博士的独立研究计划。为了满足这些需求，Molinaro博士将：1)参加耶鲁大学基因组学、生物信息学、计算机科学和分子生物学的课程；2)参加杰克逊和冷泉港实验室的哺乳动物遗传学、计算与比较基因组学、复杂性状分析等世界知名课程；3)参加科学会议和研讨会，介绍她的K22研究，建立合作，并参与有关统计基因组学当前问题的科学讨论。她提出的研究项目需要全面、积极地搜索基因组、流行病学和组织学数据，以预测感兴趣的临床结果，如复发或死亡时间。Molinaro博士建立了一个单变量方法来解决这个问题；然而，她现在需要将其扩展到现实的生物环境中。本研究项目的主要目的是：1)解释基因组变量中的缺失值；2)评价变量重要性的测度；并且，3)将这种方法扩展到包括其他统计模型，如小波和样条。这项K22资助将为Molinaro博士提供受保护的时间和资源，以完成她在癌症分子生物学方面的培训，在耶鲁大学及其他地方建立合作关系，并为科学界提供一个急需的工具，将基因组数据与临床结果联系起来。相关性：Molinaro博士的研究结合了基因组、组织学和流行病学信息，以预测临床结果，如疾病进展时间。正是这样的方法将更清楚地了解癌变的复杂性，并允许更有针对性的癌症预防和控制方法。

DESCRIPTION (provided by applicant): Dr. Annette Molinaro is an Assistant Professor in the Division of Biostatistics in the Department of Epidemiology and Public Health at Yale University School of Medicine. Prior to arriving at Yale, Dr. Molinaro was a Cancer Prevention Fellow at the National Cancer Institute. Her long term career goal is to develop statistical and computational methods which elucidate mechanisms of cancer pathogenesis to be used for the purposes of cancer prevention, diagnosis, and treatment. To reach this goal she has outlined two areas which are in need of enhancement: 1) her knowledge of functional genomics specifically related to carcinogenesis; and, 2) her proficiency in computer programming for the purposes of searching for and extracting pertinent information from vast data structures. A comprehensive understanding of the biological mechanisms behind carcinogenesis as well as the advanced computational skills necessary to implement novel statistical methods will propel Dr. Molinaro's independent research program. To meet these needs, Dr. Molinaro will: 1) attend classes at Yale University in genomics, bioinformatics, computer science, and molecular biology; 2) participate in world renowned courses at the Jackson and Cold Spring Harbor Laboratories in mammalian genetics, computational and comparative genomics, and complex trait analysis; and, 3) attend scientific meetings and workshops to present her K22 research, build collaborations, and engage in scientific discussion on current issues concerning statistical genomics. Her proposed research project entails a comprehensive, aggressive search of genomic, epidemiologic, and histologic data for the purposes of predicting a clinical outcome of interest, such as time to recurrence or death. Dr. Molinaro has established a univariate approach to this problem; however, she now needs to expand this to a realistic biological setting. The primary aims of this research project are: 1) to account for missing values in the genomic variables; 2) evaluate measures of variable importance; and, 3) extend this approach to encompass other statistical models such as wavelets and splines. This K22 grant will enable Dr. Molinaro the protected time and resources to accomplish her training in the molecular biology of cancer, establish collaborations at Yale University and beyond, and provide the scientific community with a much needed tool for associating genomic data with clinical outcomes. Relevance: Dr. Molinaro's research incorporates genomic, histological, and epidemiological information in order to predict a clinical outcome, such as time to disease progression. It is methods such as this that will provide greater clarity within the complexity of carcinogenesis and allow for more targeted methods of cancer prevention and control.