Bridge2AI: Salutogenesis Data Generation Project

人工智能准备和公平的糖尿病洞察地图集（AI-READI）项目是NIH共同基金Bridge2AI计划中的数据生成项目之一。该项目旨在创建一个基于伦理的旗舰数据集，使未来几代人工智能/机器学习（AI/ML）研究能够为2型糖尿病（T2DM）提供关键见解，包括恢复健康的有益途径。由于缺乏设计良好、高质量、大型和包容性的多模式数据集，理解和影响复杂的多器官疾病（如T2DM）病程的能力受到限制。研究小组的目标是收集4000多人的横断面数据集，以及来自美国10%的研究队列的纵向数据。将招募相同数量的黑人、西班牙裔/拉丁裔、亚洲人和白人参与者，研究队列将根据糖尿病疾病阶段进行平衡。数据收集将专门设计用于允许下游伪时间流形分析，这是一种通过收集和学习来自不同疾病严重程度（正常到胰岛素依赖型T2DM）参与者的复杂、多模式数据来预测疾病轨迹的方法。该项目的长期目标是开发糖尿病的基础数据集，不受现有分类标准或偏见的影响，可用于重建T2DM发展和健康逆转的时间图谱（即健康发生）。来自8个机构的6个跨学科项目模块将共同开发这个旗舰数据集。所有数据将为下游AI/ML研究进行优化，并公开提供。该项目还将为道德和公平的研究创建一个路线图，重点关注研究参与者和研究过程各个阶段（研究设计和数据收集、策展、分析、共享和协作）所涉及的劳动力的多样性。AI-READI项目还将参与部落协商，以解决参与的障碍和促进因素，以合乎道德和尊重的方式在美洲原住民群体中收集类似数据。具体目标包括：1)根据可查找、可访问、可互操作、可重用（FAIR）数据原则收集和共享AI/ML研究的数据集；2)创建开发多样化和代表性数据集的模型；3)通过招聘和培训具有不同背景的人员来增加AI/ML研究的可访问性和质量。

The Artificial Intelligence Ready and Equitable Atlas for Diabetes Insights (AI-READI) project is one of the data generation projects in the NIH Common Fund’s Bridge2AI program. The project seeks to create a flagship ethically-sourced dataset to enable future generations of artificial intelligence/machine learning (AI/ML) research to provide critical insights into type 2 diabetes mellitus (T2DM), including salutogenic pathways to return to health. The ability to understand and affect the course of complex, multi-organ diseases such as T2DM has been limited by a lack of well-designed, high quality, large, and inclusive multimodal datasets. The team of investigators will aim to collect a cross-sectional dataset of 4,000+ people and longitudinal data from 10% of the study cohort across the US. An equal number of Black, Hispanic/LatinX, Asian, and White participants will be recruited and the study cohort will be balanced for diabetes disease stage. Data collection will be specifically designed to permit downstream pseudotime manifold analysis, an approach used to predict disease trajectories by collecting and learning from complex, multimodal data from participants with differing disease severity (normal to insulin-dependent T2DM). The long-term objective for this project is to develop a foundational dataset in diabetes, agnostic to existing classification criteria or biases, which can be used to reconstruct a temporal atlas of T2DM development and reversal towards health (i.e., salutogenesis). Six cross-disciplinary project modules involving teams located across eight institutions will work together to develop this flagship dataset. All data will be optimized for downstream AI/ML research and made publicly available. This project will also create a roadmap for ethical and equitable research that focuses on the diversity of the research participants and the workforce involved at all stages of the research process (study design and data collection, curation, analysis, and sharing and collaboration). The AI-READI project will also engage in a tribal consultation to address barriers and facilitators of participation with the goal of collecting similar data within a Native American cohort in an ethical and respectful manner. Specific aims include 1) Collect and share the dataset for AI/ML research according to the Findable, Accessible, Interoperable, Reusable (FAIR) data principles, 2) Create a model for developing diverse and representative datasets, and 3) Increase access to and quality of AI/ML research by recruiting and training personnel with diverse backgrounds.