A-STOR Cancer Clinical Trial Artificial Intelligence & Machine Learning Readiness

项目总结 高维分子图谱，包括基因组测序和数字病理学，已经带来了革命性的变化 我们对癌症的了解，并导致了多个新的靶点和生物标志物的识别。然而， 障碍阻碍了这些数据的全部潜力的释放，因为大多数来自国家数据中心的数据 癌症研究所的国家临床试验网络(NCTN)临床试验是分散的，数据存储在 在当地进行的各种现场和分析。为联盟NCTN合作克服这些障碍 团体试验，2020年我们建立了A-STOR：联盟标准化的翻译OMICS资源，一个安全的， 符合HIPAA标准的来自Alliance的原始和经过处理的基因组和病理图像数据的集中式存储库 临床试验，与标准化的基因组数据协调，以促进交叉研究分析。在本建议书中 (联盟U10拨款的补充)，我们的目标是实现A-STOR在人工智能方面的潜力 (AI)/机器学习(ML)建模，通过快速扩展我们现有的基础设施来实现：托管来自 十几项现有或正在进行的Alliance/NCTN临床试验(目标1)，为基于云的AI/ML优化基础设施 分析包括与AI/ML应用程序的接触和反馈(目标1-2)，确保在 开发统一的临床和不良事件(AE)数据字典，以促进临床数据协调(AIM 2)，并完成一个已经批准的基于混合多模式ML的预测器作为先导研究(目标3)。这个 这些目标的影响将共同利用联盟A-STOR联合主席丹尼尔·斯托弗博士的专业知识 (临床信息学、基因组学)和James Blachly博士(计算生物学、计算机科学)，以优化 基础基础设施、临床/AE协调以促进数据重用作为AI/ML工作的一部分，以及云- 基于AI/ML对Alliance NCTN癌症临床试验生成的数据进行计算优化。

PROJECT SUMMARY High-dimensional molecular profiling, including genomic sequencing and digital pathology, has revolutionized our understanding of cancer and has led to the identification of multiple new targets and biomarkers. However, barriers impede unlocking the full potential of these data because most ‘omic data generated from National Cancer Institute’s National Clinical Trials Network (NCTN) clinical trials are decentralized, with data housed at a variety of sites and analyses taking place locally. To overcome these barriers for Alliance NCTN cooperative group trials, in 2020 we established A-STOR: Alliance Standardized Translational Omics Resource, a secure, HIPAA-compliant centralized repository of raw, and processed genomic and pathology image data from Alliance clinical trials, with standardized genomic data harmonization to facilitate cross-study analyses. In this proposal (a supplement to the Alliance U10 grant), we aim to realize the potential of A-STOR for artificial intelligence (AI)/machine learning (ML) modeling by rapidly expanding our existing infrastructure to: host data from over a dozen existing or ongoing Alliance/NCTN clinical trials (Aim 1), optimize infrastructure for cloud-based AI/ML analyses including engagement with and feedback from AI/ML applications (Aim 1-2), ensure FAIR principles in developing a unified clinical and adverse event (AE) data dictionary to facilitate clinical data harmonization (Aim 2), and complete an already-approved pooled multi-modal ML-based predictor as a pilot study (Aim 3). The impact of these aims will collectively capitalize on the expertise of Alliance A-STOR Co-Chairs Dr. Daniel Stover (clinical informatics, genomics) and Dr. James Blachly (computational biology, computer science), to optimize foundational infrastructure, clinical/AE harmonization to facilitate data reuse as part of AI/ML efforts, and cloud- based AI/ML compute optimization for data generated through Alliance NCTN cancer clinical trials.