NHLBI ENTERPRISE ARCHITECTURE AND CYBER SECURITY SUPPORT FOR DATA SCIENCE PROGRAMS

在过去的几十年里，NHLBI通过支持创建许多与心脏，肺，血液和睡眠表型相关的观察，流行病学和纵向数据集，投资于创建重要的研究和开发资源，目的是发现可用于改善患者结局的新型治疗，干预或预防策略的见解。新技术和有利的成本轨迹使这些研究参与者的详细特征成为可能，包括全基因组测序（和其他组学）和数十万参与者的成像。这些数据与动物和细胞模型相结合，增加了数据驱动的转化科学的机会。我们已全面进入“大数据”竞技场，在这一领域，我们既遇到前所未有的机遇，也面临前所未有的挑战。目前用于分析和组合这些数据集的范例受到实际和概念约束的限制。在这里，我们介绍NHLBI BioData Catalyst，这是一个由平台、工具和数据组成的新型生态系统，旨在实现和加速科学发现。 集中式架构并不是BDC团队的核心关注点，在生态系统中使用非标准和不同的架构存在挑战。有机会定义和实施标准，并在计划中执行架构服务，以加强后端，同时改善BDC上的用户体验。此外，还需要网络安全支持，因为该计划使系统与定制的风险管理框架（RMF）保持一致。 美国国立卫生研究院的人工智能/机器学习（AI/ML）联盟促进健康公平和研究多样性（AIM-AHEAD）计划建立合作伙伴关系并资助项目，以建立人工智能研究能力。具体而言，AIM-AHEAD旨在提高代表性不足的社区的能力，并有助于推进公平驱动的AI/ML方法，最大限度地减少医疗保健领域的偏见，同时解决国家健康差距。AIM-AHEAD数据基础设施方法利用更广泛的解决方案来满足参与该计划的社区的需求。AIM-AHEAD利用云平台集成数据存储、计算周期、安全性，通常还包括地理上分布的用户和组的分析工具。当数据无法汇集并且也正在程序中探索时，分布式或联合学习方法更合适。

Over the last several decades, NHLBI has invested in creating a significant resource for research and development by supporting the creation of many observational, epidemiological, and longitudinal datasets related to heart, lung, blood and sleep phenotypes, with the aim of uncovering insights that may be leveraged toward novel therapeutic, interventional, or preventive strategies resulting in improved patient outcomes. New technologies and favorable cost trajectories have enabled detailed characterization of these study participants including whole genome sequencing (and other omics) and imaging on hundreds of thousands of participants. Together this data coupled with animal and cellular models increase opportunities for data-driven translational science. We have fully entered the “Big Data” arena, in which we encounter both unprecedented opportunities as well as challenges. Current paradigms for analyzing and combining these datasets are limited by both practical and conceptual constraints. Here we introduce the NHLBI BioData Catalyst, a novel ecosystem of platforms, tools, and data to enable and accelerate scientific discovery. A centralized architecture has not been a core focus on the BDC teams and there are challenges on having non-standard and disparate architectures work within the ecosystem. There is an opportunity to define and implement standards and perform architecture services within the program to strengthen the back-end while improving the user experience on BDC. Additionally there is a need for cybersecurity support as the program aligns systems to the tailored risk management framework (RMF). The NIH’s Artificial Intelligence/Machine Learning (AI/ML) Consortium to Advance Health Equity and Research Diversity (AIM-AHEAD) Program establishes partnerships and funds projects to build AI research capacity. Specifically, AIM-AHEAD seeks to improve capacity within underrepresented communities and contributes to advancing equity-driven AI/ML approaches that minimize bias in the healthcare space while addressing national health disparities. The AIM-AHEAD data infrastructure approach utilizes a wider variety of solutions to meet the needs of the communities participating in the program. AIM-AHEAD utilizes Cloud platforms to integrate data storage, computing cycles, security, and, often, analysis tools for geographically distributed users and groups. Distributed or federated learning approaches are more appropriate when data cannot be pooled and are also being explored in the program.