InstaDeepAI/IDP-Euka-90
收藏资源简介:
--- dataset_name: IDP-Euka-90 pretty_name: IDP-Euka-90 --- # IDP-Euka-90 **IDP-Euka-90** is a collection of eukaryotic protein sequences curated for representation learning and downstream analysis of **intrinsically disordered proteins/regions (IDPs/IDRs)**. These sequences were obtained by running [Metapredict V3](https://pubmed.ncbi.nlm.nih.gov/34480923/) on all the available eukaryota proteomes from [UniProt](https://www.uniprot.org), which extracted all the IDP regions in each proteome. The sequences we subsequently clustered at 90% with [mmseqs2](https://github.com/soedinglab/MMseqs2) to remove all near duplicates. The HF dataset is distributed as a train/val split and the backup csv has a **`split` column** prepared. --- ## Contents - **Columns** - `sequence` — protein amino-acid sequence (single-letter codes) - **Format** - Hosted on the Hugging Face Hub as an Arrow/CSV-backed dataset. --- ## Quick start ```python from datasets import load_dataset repo_id = "InstaDeepAI/IDP-Euka-90" ds = load_dataset(repo_id) print(ds) print(ds.features)
--- dataset_name: IDP-Euka-90 pretty_name: IDP-Euka-90 --- # IDP-Euka-90 **IDP-Euka-90** 是一款专为内在无序蛋白质/区域(intrinsically disordered proteins/regions,IDPs/IDRs)的表征学习与下游分析任务构建的真核生物蛋白质序列精选集合。 这些序列通过将Metapredict V3工具应用于来自通用蛋白质资源库(UniProt)的所有可用真核生物蛋白质组得到,该工具可从每个蛋白质组中提取所有内在无序蛋白质区域。随后,我们使用mmseqs2工具以90%的序列相似度对序列进行聚类,以去除所有近重复序列。 该Hugging Face(HF)数据集以训练集/验证集划分(train/val split)的形式发布,其备份CSV文件中已预先设置好**`split` 列**。 --- ## 数据集内容 - **字段说明** - `sequence` — 蛋白质氨基酸序列(采用单字母编码格式) - **存储格式** - 以基于Arrow/CSV格式的数据集形式托管于Hugging Face Hub平台。 --- ## 快速上手示例 python from datasets import load_dataset repo_id = "InstaDeepAI/IDP-Euka-90" ds = load_dataset(repo_id) print(ds) print(ds.features)




