Presentation Information
[8p-P11-1]Graph Convolutional Network for Predicting Atomic Charges of Various Organic
〇(B)Yugo Homma1, Yu Homma1, Hiroyuki Matsui1 (1.ROEL, Yamagata Univ.)
Keywords:
machine learning,atomic charges prediction,graph convolutional network
In recent years, there have been advances in high-precision atomic charge prediction using machine learning. However, most of these methods rely on datasets centered on drug molecules generated on computers. Therefore, their applicability to conjugated and aromatic compounds, which are important in organic semiconductor research, remains unclear. In this study, we used the Cambridge Structural Database (CSD), which contains a large number of conjugated and aromatic compounds, to develop a new, highly versatile atomic charge prediction model using graph convolutional networks (GCN).