ConvPred_data

The file <strong>"USPTO_DUT_Wang.xlsx</strong>" contains the reactions that are extracted from the USPTO 1976-2016 databse. It only contains the reactions that are in the form of A+B&gt;C to make sure all the reactions are compatible with our convolutional algorithm. We used sigma-profiles and ECFP4 to train two different models. They are available in the "<strong>.npy</strong>" files where molecules and their feature vectors are stored in the form of key and value pairs (i.e., python dict). <br> The "<strong>VRF(sigma).7z</strong>" contains the python script used to generate certain VRF(sigma) for chemical reactions. <br> The "<strong>Scripts for extracting reactions and templates.7z</strong>" contains the python scripts used to extract certain types of reactions and templates from the USPTO 1976-2016 database.文件**USPTO_DUT_Wang.xlsx**收录了从USPTO 1976-2016数据库中提取的化学反应数据。该文件仅收录形式为A+B→C的化学反应，以确保所有数据均可适配本研究的卷积算法。 本研究采用σ特征谱（sigma-profiles）与扩展连接性指纹4（ECFP4）训练了两款差异化模型，相关模型存储于**.npy**格式文件中，该类文件以键值对（即Python字典）的形式存储分子及其特征向量。 **VRF(sigma).7z**压缩包内含用于生成特定化学反应VRF(sigma)的Python脚本。 **Scripts for extracting reactions and templates.7z**压缩包内含用于从USPTO 1976-2016数据库中提取特定类型反应与反应模板的Python脚本。