A DATA SCIENCE APPROACH TO AIR TOXICS AND CHILDREN'S ENVIRONMENTAL HEALTH

项目摘要：本提案旨在描述产前暴露与 通过有效利用大型公共设施， 健康数据。在多学科顾问的指导下，候选人将发展数据科学技能， 机器学习和先进的生物统计学来补充她在流行病学方法方面的训练。这将 使她能够在职业生涯中取得进步，并推进对综合环境暴露的研究， 儿童健康。先前的研究发现，产前暴露于单一空气污染物 和儿童的认知健康，但缺乏调查多种污染物综合影响的能力， 包括在实验中观察到的污染物之间的协同/拮抗相互作用， 问题研究了解综合照射的影响是国家放射学研究所的一个战略目标。 环境健康科学，环境健康领域正在从单一污染物 更全面的模式，如麻烦。确定内部的关联和相互作用 高维曝光数据的上下文提出了计算挑战。来自域的方法 例如数据科学，包括机器学习方法，可以被纳入流行病学工具箱， 用于解决环境混合物和多重暴露问题。这个职业发展奖的目标是 在大数据科学的交叉点推进候选人进入独立的研究生涯， 儿童的环境健康。通过正式的课程，定向学习和实地轮换， 候选人将获得数据科学，机器学习和高级生物统计学方面的技能。导师、顾问和 顾问公司是根据其互补的专业知识、相关的研究经验和 指导能力。拟议的研究将利用从培训计划中获得的技能并加以应用 描述产前暴露于可解释的空气毒性组合和第三代 年级标准化考试成绩，一个以学校为基础的认知成果的衡量标准。出生时的居住地 用于将空气有毒物质（空气污染物的一个子集）的数据与公共卫生的行政数据联系起来 1994-1998年出生于纽约市的大约22万名儿童的登记和教育数据。的 候选人将开发和验证假设生成的两阶段方法，然后有针对性地 分析，以确定与儿童测试分数相关的空气毒性组合 高维暴露数据（目标1）。使用成熟的流行病学方法进行目标分析 进行影响估计和空气毒物之间相互作用的评估。（目标2）潜力 空气中有毒物质和考试成绩之间的关系的调解人，然后可以确定使用统计 调解和数据科学方法。(Aim 3）完成这些目标将使候选人处于独特的地位 开展未来关于综合环境暴露和儿童健康的研究。

PROJECT SUMMARY: This proposal aims to characterize the associations between prenatal exposure to interpretable combinations of air toxics and children’s cognitive health through the efficient use of big public health data. With guidance from multidisciplinary advisors, the candidate will develop skills in data science, machine learning and advanced biostatistics to supplement her training in epidemiologic methods. This will allow her to progress in her career and advance research on combined environmental exposures and children’s health. Previous research has found associations between prenatal exposure to single air pollutants and children’s cognitive health but has lacked the ability to investigate combined impacts of multiple pollutants, including the synergistic/antagonistic interactions between pollutants that have been observed in experimental studies. Understanding the effects of combined exposures is a strategic goal of the National Institute of Environmental Health Sciences, and the field of environmental health is transitioning from single-pollutant approaches to more holistic paradigms, such as the exposome. Identifying associations and interactions within the context of high-dimensional exposure data presents a computational challenge. Methods from domains such as data science, including machine learning methods, can be incorporated into the epidemiologic toolbox for addressing environmental mixtures and multiple exposures. The goal of this Career Development Award is to advance the candidate into an independent research career at the intersection of big data science and children’s environmental health. Through formal coursework, directed learning and field rotations, the candidate will gain skills in data science, machine learning and advanced biostatistics. Mentors, advisors and consultants have been selected for their complementary expertise, relevant research experience and mentoring abilities. The proposed research will leverage the skills gained from the training plan and apply them to characterize associations between prenatal exposure to interpretable combinations of air toxics and 3rd grade standardized test scores, a school-based measure of cognitive outcomes. Residence at birth will be used to link data on air toxics, a subset of air pollutants, to an administrative data linkage of public health registries and education data for approximately 220,000 children born in New York City from 1994-1998. The candidate will develop and validate a two-stage approach of hypotheses generation followed by targeted analyses in order to identify combinations of air toxics associated with children’s test scores within the context of high-dimensional exposure data (Aim 1). Targeted analyses using well-established epidemiologic methods for effect estimation and assessment of interaction between air toxics will be performed. (Aim2) Potential mediators of the relationship between air toxics and test scores can then be identified using statistical mediation and data science approaches. (Aim 3) Completion of these aims will uniquely position the candidate to conduct future research on combined environmental exposures and children’s health.