基于真实体外药物反应的900种癌症细胞系数据集

本数据集由德国癌症研究中心等机构创建，包含约900种癌症细胞系的体外药物反应数据。数据集详细记录了每种细胞对7种临床常用药物的完整反应，支持计算决策策略的生物学合理反事实。创建过程中，数据预处理包括对IC50值的日志转换和标准化，以及使用UMAP技术将18523个细胞特异性特征降维至20维。数据集旨在为精准肿瘤学中的上下文盗贼算法提供基准，解决个性化治疗分配的效率问题，特别是在考虑基因组信息的情况下。

This dataset was developed by the German Cancer Research Center and other collaborating institutions, and comprises in vitro drug response data for approximately 900 cancer cell lines. It comprehensively documents the full response profiles of each cell line against 7 clinically commonly used drugs, enabling biologically plausible counterfactual analyses for computational decision-making strategies. During its construction, data preprocessing workflows included log-transformation and standardization of IC50 values, as well as dimensionality reduction of 18,523 cell-line-specific features down to 20 dimensions using UMAP. This dataset is intended to serve as a benchmark for contextual bandit algorithms in precision oncology, addressing the efficiency issues of personalized treatment allocation, particularly when genomic information is incorporated.