MolTextNet

该数据集名为MolTextNet，包含了250万个高质量的分子的文本配对样本，旨在支持多种分子学习任务。这些样本配备了由GPT-4生成的结构化描述，融合了分子的结构特征、物理化学性质、生物活性数据以及合成复杂性。该数据集的规模达到了250万分子与文本的配对，可应用于多模态分子学习、性质预测和结构检索等多种任务。

This dataset, designated as MolTextNet, comprises 2.5 million high-quality molecular-text paired samples, intended to support a wide range of molecular learning tasks. Each sample is furnished with structured descriptions generated by GPT-4, which integrate molecular structural features, physicochemical properties, bioactivity data, and synthetic complexity. Boasting a total of 2.5 million molecular-text pairs, this dataset is applicable to diverse tasks including multimodal molecular learning, property prediction, and structure retrieval.