AIM-Harvard/TrialPanorama-filtered

TrialPanorama-filtered是TrialPanorama/TrialPanorama-database的过滤版本，专注于临床试验数据。该数据集包含多种配置，如研究、条件、药物、处置、结果、生物标志物、终点、关系和药物作用机制。过滤过程移除了与特定任务CSV中ID重叠的study_id行。这是一个大规模、结构化的数据集，旨在支持AI驱动的临床试验工作流程任务。它汇总了来自ClinicalTrials和PubMed的超过100万条临床试验记录，捕获了试验设置、干预措施、条件、生物标志物、结果以及与生物医学本体（如DrugBank、MedDRA）的链接等标准化元素。数据集支持系统评价和试验设计领域的8个基准任务。

TrialPanorama-filtered is a filtered version of the TrialPanorama/TrialPanorama-database, focusing on clinical trial data. The dataset includes various configurations such as studies, conditions, drugs, disposition, outcomes, results, biomarkers, endpoints, relations, and drug_moa. The filtering process removes rows where study_id overlaps with IDs present in specific task CSVs. This is a large-scale, structured dataset designed to support AI-driven tasks in clinical trial workflows. It aggregates over 1M clinical trial records from ClinicalTrials and PubMed, capturing standardized elements such as trial setups, interventions, conditions, biomarkers, outcomes, and links to biomedical ontologies (e.g., DrugBank, MedDRA). The dataset supports a suite of 8 benchmark tasks across systematic review and trial design domains.