COMPASS ICI response prediction dataset

This record contains the <b>Immunotherapy Response Patient (ITRP)</b> dataset used for downstream fine-tuning and validation of the <b>COMPASS</b> foundation model.The dataset aggregates <b>1,133 patients</b> from <b>16 published immunotherapy cohorts (</b>including RNA-seq TPM matrices, clinical annotations, and response labels.<b>)</b>, spanning melanoma, bladder, renal, lung, gastric, and brain cancers. Each patient received immune checkpoint inhibitors (anti–PD-1, anti–PD-L1, or anti–CTLA-4).<b>Contents</b>The ZIP archive includes two pandas pickle files:<b>ITRP.TPM.TABLE</b> — gene-level RNA-seq TPM matrix across 1,133 patients.<b>ITRP.PATIENT.TABLE</b> — clinical metadata table containing cancer type, treatment, and response labels (responder vs. non-responder).These files were preprocessed and harmonized using the official COMPASS data-processing pipeline to ensure consistent gene naming, normalization, and cohort integration.<br>Documentation and preprocessing scripts are available at:<br>🔗 https://www.immuno-compass.com/help/index.html#datasets<br>🔗 https://github.com/mims-harvard/COMPASS-web/tree/main/TCGA_dataset_processing<b>Data access</b>Due to data-sharing and privacy regulations, the underlying patient-level data were originally obtained from controlled repositories such as <b>EGA</b>, <b>dbGaP</b>, and <b>ENA</b>.<br>Redistribution is restricted; users should request access from the corresponding repositories or original publications.

本数据集为**免疫治疗应答患者（Immunotherapy Response Patient, ITRP）**数据集，用于**COMPASS**基础模型的下游微调与验证。该数据集整合了来自16项已发表免疫治疗队列的1133名患者数据，涵盖RNA-seq TPM矩阵、临床注释信息及应答标签，涉及黑色素瘤、膀胱癌、肾癌、肺癌、胃癌及脑癌。所有入组患者均接受了免疫检查点抑制剂治疗（抗PD-1、抗PD-L1或抗CTLA-4）。

**数据集内容**
该ZIP压缩包包含两个pandas pickle格式文件：
- **ITRP.TPM.TABLE**：覆盖1133名患者的基因级RNA-seq TPM表达矩阵。
- **ITRP.PATIENT.TABLE**：临床元数据表，包含癌症类型、治疗方案及应答标签（应答者vs. 无应答者）。

所有文件均通过官方COMPASS数据处理管线完成预处理与标准化，以确保基因命名、归一化流程及队列整合的一致性。

相关文档与预处理脚本可通过以下链接获取：
🔗 https://www.immuno-compass.com/help/index.html#datasets
🔗 https://github.com/mims-harvard/COMPASS-web/tree/main/TCGA_dataset_processing

**数据获取说明**
受数据共享与隐私监管要求限制，本研究的患者级原始数据最初从**EGA（European Genome-phenome Archive）**、**dbGaP（Database of Genotypes and Phenotypes）**及**ENA（European Nucleotide Archive）**等受控数据库获取。数据集的重新分发受到严格限制，用户需向对应数据库或原始文献申请访问权限。