jspector792/Geneformer_NetworkMedicine

该数据集包含了用于生成论文《Transformers Enhance the Predictive Power of Network Medicine》中图形的聚合权重矩阵。数据集结构分为aggregated_attentions和aggregated_embeddings两个主要部分，每个部分包含dcm_samples和genecorpus两个子集，分别对应fine_tuned和pretrained模型的数据。每个文件夹中包含三个文件：counts.pkl记录基因共现次数，vals.pkl记录基因间的聚合权重，gene_dict.pkl为基因索引字典。

This dataset contains aggregated weight matrices used to generate figures in the manuscript Transformers Enhance the Predictive Power of Network Medicine. The dataset structure is divided into two main parts: aggregated_attentions and aggregated_embeddings, each containing subsets for dcm_samples and genecorpus, corresponding to fine_tuned and pretrained models. Each folder contains three files: counts.pkl records gene co-occurrence counts, vals.pkl records the aggregated weights between genes, and gene_dict.pkl is a dictionary specifying the index of each gene.