lamm-mit/silkome-masp
收藏资源简介:
Silkome MaSp数据集是一个用于蜘蛛丝蛋白研究的专业数据集,专注于主要壶腹丝蛋白(MaSp)序列与纤维级机械属性之间的关联。该数据集从完整的Spider Silkome资源(lamm-mit/silkome-full)中筛选出“category1”为MaSp、MaSp1、MaSp2、MaSp2B、MaSp3或MaSp3B的条目,共包含1,033行数据,涉及1,028个唯一蛋白质序列和233个唯一纤维/属性标识符(idv)。数据集中每个行提供蛋白质级丝蛋白序列(氨基酸序列长度范围从115到1,854个氨基酸)以及对应的纤维级机械测量值,包括韧性(toughness)、杨氏模量(E)、强度(strength)和应变(strain)等核心属性及其标准偏差(SD)。数据集提供了完整分割(full)、训练分割(train)和测试分割(test),其中训练/测试分割基于idv分组进行确定性随机划分,确保无重叠,以支持序列到属性预测、条件序列生成和蛋白质设计等研究。该数据集旨在用于蜘蛛丝蛋白语言建模、材料科学工作流程以及MaSp序列基序与机械性能分析,但需注意机械属性是纤维级测量,而序列是蛋白质级,因此属于弱监督资源,适用于生成建模和设计研究。
The Silkome MaSp dataset is a specialized resource for spider silk protein research, focusing on the association between major ampullate spidroin (MaSp) sequences and fiber-level mechanical properties. It is curated from the full Spider Silkome resource (lamm-mit/silkome-full) by selecting rows where category1 is one of MaSp, MaSp1, MaSp2, MaSp2B, MaSp3, or MaSp3B, resulting in 1,033 rows with 1,028 unique protein sequences and 233 unique fiber/property identifiers (idv). Each row includes a protein-level spidroin sequence (amino acid sequence length ranging from 115 to 1,854 residues) paired with corresponding fiber-level mechanical measurements, such as core properties like toughness, Youngs modulus (E), strength, strain, and their standard deviations (SD). The dataset provides full, train, and test splits, with the train/test splits deterministically randomized and grouped by idv to ensure no overlap, supporting research in sequence-to-property prediction, conditional sequence generation, and protein design. It is intended for applications in spider silk protein language modeling, materials science workflows, and analysis of MaSp sequence motifs and mechanical performance, but note that the mechanical targets are fiber-level measurements while sequences are protein-level, making it a weakly supervised resource suitable for generative modeling and design studies.




