论文部分内容阅读
为了采用计算方法研究手性化合物,为了通过结构-活性相关性研究预测与手性有关的性质,本文提出了采用σ和π的残余电负性之和作为原子属性的构象独立手性指数。该手性指数衍生于原子径向分布函数,包含分子几何和原子属性的信息,能够区分对映体。将构象独立手性指数应用于1个包含48对手性氨醇对映体的数据集,该数据集为苯甲醛与二乙基锌发生加成反应的催化剂,每个催化剂均产生特定绝对构型的反应主产物。采用相向传输神经网络建立了手性氨醇催化剂的构象独立手性指数与反应主产物绝对构型的相关性模型,得到了满意的预测结果。对于独立的测试集,90.0%的催化剂被正确地预测;对于训练集,89.5%的催化剂被正确地识别。
In order to study chiral compounds by computational methods, in order to predict chirality-related properties through structure-activity correlation studies, we propose the use of the sum of residual electronegativities of σ and π as the conformational independent chiral index of atomic properties. The chiral index is derived from the atomic radial distribution function, which contains the information of molecular geometry and atomic properties and is able to distinguish enantiomers. The conformer-independent chiral index was applied to a dataset containing 48 enantiomers of chiral aminoalcohol, which was a catalyst for the addition reaction of benzaldehyde and diethylzinc, with each catalyst producing a specific absolute configuration The main reaction product. The model of the correlation between the conformational independent chiral index of chiral aminoalcohol catalyst and the absolute configuration of the main product of the reaction was established by using the phase-transfer neural network, and the satisfactory prediction result was obtained. For the independent test set, 90.0% of the catalysts were correctly predicted; for the training set, 89.5% of the catalysts were correctly identified.