HDMA TF-MoDISco motifs per cluster

Per-cell-type motif discovery using TF-MoDISco for HDMA (Liu*, Jessa*, Kim*, Ng* et al, bioRxiv 2025). For the 189 cell types where high-quality models were trained, we provide the TF-MoDISco output as h5 files per cell type. These represent motifs learned through interpretation of the models with respect to counts. The file merged_modisco_patterns_map.tsv indicates which cell type patterns were clustered and aggregated to form compendium motifs. A detailed description of the main data types deposited on Zenodo can be found here.本研究针对HDMA数据集，采用TF-MoDISco开展每种细胞类型的基序（motif）发现工作（Liu*、Jessa*、Kim*、Ng*等研究者，bioRxiv 2025）。针对189个已完成高质量模型训练的细胞类型，我们以对应各细胞类型的h5文件形式提供TF-MoDISco的输出结果。这些文件包含通过结合计数特征对模型进行解读所学习得到的基序。文件merged_modisco_patterns_map.tsv记录了哪些细胞类型的模式被聚类并聚合以形成汇总基序。存放在Zenodo平台的主要数据类型的详细说明可在此处查阅。