PRECISE - a PErsonalized Risk Score for gastrIc CancEr

国家癌症研究所呼吁消除癌症发病率的差异 通过使用数据科学来降低死亡率。胃癌仍然是分布最不均匀的肿瘤之一， 在美国，癌症在某些民族、种族和移民群体中负担很高。识别 胃癌风险最高的个体的死亡率可能允许有针对性的风险衰减计划，并改善 健康公平。候选人和职业发展计划：我是一个委员会认证的胃肠病学家和硕士 斯坦福大学的一位受过学位训练的流行病学家，他试图利用数据科学来减少 癌症结果。根据我的培训和经验，我在胃肠癌方面有内容专业知识 诊断，流行病学原理和观察性研究设计的方法学专业知识。为了 实现我的长期目标，成为一个独立的调查员和国家领导人在癌症的差距 研究，我需要额外的定量技能（大数据分析，基于机器学习的风险预测， 使用自然语言处理的非结构化数据提取），定性技能（有效的科学 沟通，科学领导力）和专业发展。研究计划：总体研究 该提案的目的是使用真实世界的数据开发胃癌的个性化风险评分（PRECISE） 临床数据来源。我的总体假设是，通过使用先进的数据分析和深度学习， 通过这些方法，可以鉴定出胃癌风险最高的个体的高度精炼的队列。具体 本提案的目的是试图解决这一假设：（1）建立一个个性化的风险预测模型， 回归，（2）使用机器学习算法构建个性化风险预测模型，以及（3） 比较电子健康记录数据中的回归和机器学习模型。实现这些目标 将产生一种新的，个性化的预测评分，这将有助于识别胃肠道疾病高风险的个体。 癌症和谁可能受益于有针对性的风险衰减计划。导师团队：为了实现这些目标 我组建了一个世界级的导师团队，他们拥有流行病学和健康差异方面的专业知识， 研究（Latha Palaniappan，主要导师）、EHR数据中的机器学习和自然语言处理 (Tina Hernandel-Boussard，共同导师）和胃癌筛查和预防（Joo Ha Hwang，共同导师）。 环境和机构承诺：斯坦福大学是世界领先的临床 信息学、流行病学和卫生服务研究。我就能得到一个独一无二的数据核心 世界上最广泛和最强大的精选临床数据集之一。我的导师团队 我致力于确保提案的成功，并将我培养成一名独立调查员 竞争力的R级赠款。

The National Cancer Institute has called for eliminating disparities in cancer morbidity and mortality through the use of Data Science. Gastric cancer remains one of the most unequally distributed cancers in the United States, with high burden among certain ethnic, racial, and immigrant groups. Identification of individuals at greatest risk for gastric cancer may allow for targeted risk attenuation programs, and improve health equity. Candidate and Career Development Plan: I am a board-certified Gastroenterologist and Master’s degree-trained epidemiologist at Stanford University who seeks to use data science to reduce disparities in cancer outcomes. Based on my training and experience, I have content expertise in gastrointestinal cancer diagnosis, and methodologic expertise in epidemiologic principles and observational study design. In order to achieve my long-term goal of becoming an independent investigator and national leader in cancer disparities research, I require additional quantitative skills (large data analytics, machine learning-based risk prediction, unstructured data extraction using natural language processing), qualitative skills (effective scientific communication, scientific leadership), and professional development. Research Plan: The overarching research aim of this proposal is to develop a PErsonalized Risk Score for gastrIc CancEr (PRECISE) using real-world clinical data sources. My overall hypothesis is that through use of advanced data analytics and deep learning methods, a highly-refined cohort of individuals at highest risk for gastric cancer can be identified. The Specific Aims of this proposal seek to address this hypothesis: (1) to build a personalized risk prediction model using regression, (2) to build a personalized risk prediction model using machine learning algorithms, and (3) to compare regression and machine learning models in electronic health records data. Achievement of these aims will produce a novel, personalized prediction score which will help identify individuals at high risk for gastric cancer and who may benefit from targeted risk attenuation programs. Mentorship Team: To achieve these Aims, I have assembled a world class mentorship team with expertise in epidemiology and health disparities research (Latha Palaniappan, primary mentor), machine learning and natural language processing in EHR data (Tina Hernandez-Boussard, co-mentor), and gastric cancer screening and prevention (Joo Ha Hwang, co-mentor). Environment and Institutional Commitment: Stanford University is a world leader in clinical informatics, epidemiology, and health services research. I will have access to a unique data core, which contains one of the most extensive and robust collections of curated clinical data in the world. My mentorship team is committed to ensuring the success of the proposal, and in developing me to become an independent investigator competitive for R-level grants.