ConvergeBio/virtual-cell-patient-example

Virtual Cell — Patient Example Dataset是一个最小样本数据集，用于验证数据格式并使用ConvergeBio/virtual-cell-patient模型进行快速端到端检查。该数据集不适用于训练或评估，仅包含少量患者数据，不代表真实分布。数据集内容来源于一项公开的1型糖尿病单细胞RNA测序研究（GSE148073），并已预处理为模型输入格式作为最小工作示例。数据集包含训练集（8个患者，40行）和验证集（3个患者，15行），每行代表一个患者的增强视图（每个患者有5个增强）。数据模式包括input_ids（基因表达矩阵）、attention_mask（细胞掩码）、labels（类别索引）、entity_id（患者标识符）和sample_id（原始样本ID）。Virtual Cell — Patient Example Dataset is a minimal sample dataset for verifying the data format and running quick end-to-end checks with ConvergeBio/virtual-cell-patient. This dataset is not intended for training or evaluation. It contains a small number of patients and is not representative of a real distribution. Derived from a public type 1 diabetes scRNA-seq study (GSE148073), preprocessed into the models input format as a minimal working example. The dataset includes train split (8 patients, 40 rows) and validation split (3 patients, 15 rows), where each row is one augmented view of a patient (5 augmentations per patient). Schema includes input_ids (gene expression matrix), attention_mask (cell mask), labels (class index), entity_id (patient identifier), and sample_id (original sample accession ID).