Awesome

Abstract

Accurate prediction of molecular properties is important for new compound design, which is a crucial step in drug discovery. In this paper, molecular graph data is utilized for property prediction based on graph convolution neural networks. In addition, a convolution spatial graph embedding layer (C-SGEL) is introduced to retain the spatial connection information on molecules. And, multiple C-SGELs are stacked to construct a convolution spatial graph embedding network (C-SGEN) for end-to-end representation learning. In order to enhance the robustness of the network, molecular fingerprints are also combined with C-SGEN to build a composite model for predicting molecular properties. Our comparative experiments have shown that our method is accurate and achieves the best results on some open benchmark datasets.

User Notification

The model is used in the company's Protac database project and is currently being integrated into the company's internal software. If need cooperation, please contact: xiaofeng@mindrank.ai