BioSample and GEO FAIR Metadata Curation with GPT-4

The repostory contains records from BioSample and GEO that relate to lung cancer, liver cancer, and ovarian cancer. he format of these files is in JSON. The raw outputs from GEO are available in the text files along with the heuristic annotations for the metadata field "tissue". They also include a curated version using carefully designed prompts that incorporate information on metadata from CEDAR using GPT-4, Medllama and Llama3.1. Data desciption The annotated-xxxx files contain the gold standards across datasets (lung, ovary, liver) and databases (BioSample, GEO). The xxxx-cancer-xxxx.txt contain the source data across the datasets and databases. The xxxx-cancer-xxxx.json contain the LLM augmented data across the datasets and databases and LLMs (GPT-4, Llama and Medllama)该仓库收录了来自生物样本数据库（BioSample）与基因表达综合数据库（GEO）中与肺癌、肝癌及卵巢癌相关的记录。所有文件均采用JSON格式存储。GEO的原始输出结果以文本文件形式提供，同时附带元数据字段"组织（tissue）"的启发式注释。此外，数据集还包含经整理优化后的版本：该版本通过精心设计的提示词，结合来自CEDAR的元数据信息，由GPT-4、Medllama及Llama3.1生成。 数据集说明 带有annotated-前缀的文件包含跨肺癌、卵巢癌、肝癌数据集以及BioSample、GEO数据库的金标准标注数据集。 命名格式为xxxx-cancer-xxxx.txt的文件包含跨各数据集与数据库的原始源数据。 命名格式为xxxx-cancer-xxxx.json的文件包含跨各数据集、数据库及大语言模型（GPT-4、Llama、Medllama）生成的经大语言模型增强的数据。