LiteFold/Pfam

Pfam是一个广泛使用的蛋白质家族和结构域注释资源，基于经过人工校正的多序列比对和隐马尔可夫模型构建。该Hugging Face数据集以标准化的JSONL格式镜像了Pfam当前版本的表格导出数据，包含行级来源信息，旨在用于大规模结构域注释查询、蛋白质家族分析、基于Pfam-A区域分配的模型训练以及与蛋白质序列或结构数据集的集成。该镜像包含两个Pfam-A表：一个小的clan映射表和一个非常大的regions表。clan表将Pfam-A家族映射到clan和家族名称，而regions表记录Pfam-A结构域在蛋白质序列上的位置。每个JSONL记录都保留了原始上游行数据（在row字段下）并添加了来源字段，以便下游用户可以追踪源文件和行索引。

Pfam is a widely used protein family and domain annotation resource built from curated multiple sequence alignments and profile hidden Markov models. This Hugging Face dataset mirrors tabular exports from the Pfam current release in a normalized JSONL format with row-level provenance. It is intended for large-scale domain annotation lookup, protein family analysis, model training over Pfam-A region assignments, and integration with protein sequence or structure datasets. This mirror contains two Pfam-A tables: a small clan mapping table and a very large regions table. The clan table maps Pfam-A families to clans and family names, while the regions table records where Pfam-A domains occur on protein sequences. Each JSONL record preserves the upstream row under `row` and adds provenance fields so downstream users can trace the source file and row index.