huggingworld/genomic-reasoning-qa

该数据集包含来自23andMe客户的631,455个SNP（单核苷酸多态性）数据，用于训练和评估一个多步推理代理系统，该系统能够解释个人基因组变异与生物医学知识数据库（如ClinVar、GWAS Catalog、gnomAD）的关系。数据集通过ClinVar API进行注释，构建了一个包含36个问答对的QA数据集，涵盖三种任务类型：变异解释、基因型解释和通路推理。每个问答对都有可验证的答案，基于ClinVar注释的真实数据。

The dataset includes 631,455 SNPs (Single Nucleotide Polymorphisms) from 23andMe client data, used to train and evaluate a multi-step reasoning agent system that interprets personal genomic variants against biomedical knowledge databases (ClinVar, GWAS Catalog, gnomAD). The dataset is annotated via ClinVar API and constructs a QA dataset with 36 question-answer pairs covering three task types: variant interpretation, genotype interpretation, and pathway reasoning. Each QA pair has verifiable answers based on ground truth from ClinVar annotations.