Development and Use of an AI/ML-Ready Dog Aging Project Dataset

摘要 狗老化项目(DAP)是一项全国性的关于遗传和环境问题的社区科学研究 伴侣犬健康衰老的决定因素。这项长期研究已经招募了30,000多人 全国范围内的参与者，目标是收集关于每个犬类参与者的丰富数据集 生活。每个参与者都提供了业主报告的健康、生活体验、认知功能、 和家庭环境，以及当可用时，兽医电子病历。本地环境数据 对于每只狗，包括空气质量，水质，天气数据，步行得分，等等。DAP将收集 10,000只狗的全基因组测序数据，其中1000多只狗，每年测量 广泛的系统生物学数据(活动、临床病理测量、代谢组、微生物组、 表观基因组和流式细胞术)。DAP是一项开放的科学研究--所有数据将提供给 世界各地的研究人员，目标是最大限度地发挥科学发现的影响 这些数据。因此，DAP数据集为那些对应用AI/ML感兴趣的人提供了一个巨大的机会 获取有趣、重要的数据集的方法。这项提案的目标有两个。首先，它将为一个数据提供资金 具有处理大型数据集以确保DAP数据最大程度符合的经验的科学家 符合AI/ML分析方法的需要。本补编资助的工作将确保工作流程 用于数据和元数据构建、用于数据的预处理、清理和过滤、用于输入 丢失数据和元数据，并维护数据文档。其次，支持AI/ML的数据至关重要， 和AI/ML方法来分析数据，DAP研究团队和更广泛的人都可以使用 社区。考虑到这一点，DAP团队正在与华盛顿大学eScience合作 研究所，全国首批数据科学研究所之一，其明确目的是提供培训和 为研究人员提供的资源，用于处理大型、复杂和有噪音的数据集，并具有丰富的专业知识 人工智能/最大限度地接近。DAP团队将与eScience Institute密切合作，设计和展示一系列 全年的网络研讨会，介绍DAP的研究人员，他们已经有了广泛的统计 在各自领域的经验，AI/ML方法的力量，以及如何在DAP数据上实施它们。 在本附录的资助期接近尾声时，eScience Institute将运行DAP AI/ML黑客 本周，对AI/ML方法感兴趣的外部研究人员将加入DAP团队。开展的工作 这里将为实施工具奠定关键基础，这些工具通过以下方式促进对DAP数据进行强大的AI/ML分析 世界各地的研究人员。

Abstract The Dog Aging Project (DAP) is a nationwide Community Science study on the genetic and environmental determinants of healthy aging in companion dogs. This long-term study has already enrolled over 30,000 participants nationwide, with the goal of collecting a rich dataset about each canine participant throughout its life. Each participant provides data from owner-reported surveys on health, life experience, cognitive function, and home environment, and when available, veterinary electronic medical records. Local environmental data for each dog include air quality, water quality, weather data, walkability scores, and more. The DAP will collect whole genome sequencing data for 10,000 dogs, and for more than 1000 of those dogs, annual measures of extensive systems biological data (actigraphy, clinicopathology measures, metabolome, microbiome, epigenome, and flow cytometry). The DAP is an Open Science study--all data will be made available to researchers around the world, with the goal of maximizing the impact of scientific discoveries that arise from these data. Thus, the DAP dataset offers a tremendous opportunity for those interested in applying AI/ML approaches to interesting, important datasets. The goals of this proposal are twofold. First, it will fund a data scientist with experience in working with large datasets to ensure that the DAP data are maximally compliant with the needs of AI/ML analytical approaches. The work funded by this Supplement will ensure that workflows are in place for data and metadata construction, for pre-processing, cleaning and filtering data, for imputing missing data and metadata, and maintaining data documentation. Second, it is critical that AI/ML-ready data, and AI/ML approaches to analyze the data, are available for the DAP research team and for the broader community. With that in mind, the DAP team is collaborating with the University of Washington eScience Institute, one of the nation's first data science institutes, whose express purpose is to provide training and resources for researchers to work with large, complex and noisy data sets, and with extensive expertise in AI/ML approaches. The DAP team will work closely with the eScience Institute to design and present a series of webinars throughout the year, introducing the DAP researchers, who already have extensive statistical experience in their own fields, to the power of AI/ML approaches and how to implement them on DAP data. Towards the end of the funding period of this Supplement, the eScience Institute will run a DAP AI/ML hack week, where outside researchers interested in AI/ML approaches will join the DAP team. The work carried out here will lay the critical groundwork to implement tools that facilitate powerful AI/ML analyses of DAP data by researchers around the world.