TextOmics
收藏资源简介:
TextOmics是一个开创性的基准数据集,它建立了组学表达与分子文本描述之间的一对一对应关系。TextOmics提供了一种异构数据集,通过表示对齐来促进分子生成。该数据集包含了反映生物环境、传递语义知识和编码化学结构的SELFIES表示等三种异构数据。SELFIES作为一种桥梁表示,有效地将组学表达和文本描述联系起来。TextOmics数据集的构建过程涉及将分子暴露于特定的细胞环境,以获得相应的组学表达,并通过BioT5将SELFIES表示转换为相应的文本描述。TextOmics旨在解决药物发现中缺乏异构数据和统一框架的问题,用于生成具有治疗潜力的类似命中分子。
TextOmics is a groundbreaking benchmark dataset that establishes one-to-one correspondence between omics expressions and molecular text descriptions. TextOmics provides a heterogeneous dataset that facilitates molecular generation via representation alignment. This dataset includes three types of heterogeneous data: SELFIES representations that reflect biological contexts, convey semantic knowledge, and encode chemical structures. As a bridge representation, SELFIES effectively links omics expressions and text descriptions. The construction process of the TextOmics dataset involves exposing molecules to specific cellular environments to obtain corresponding omics expressions, and converting SELFIES representations into corresponding text descriptions via BioT5. TextOmics aims to address the lack of heterogeneous data and unified frameworks in drug discovery, and supports the generation of hit-like molecules with therapeutic potential.




