Augmenting optimization-based molecular design with graph neural networks

Shiqiang Zhang, Juan S. Campos, Christian Feldmann, Frederik Sandfort, Miriam Mathea, Ruth Misener

Abstract: Highlights•We formulate graph neural networks with mixed-integer programming and break symmetry.•We integrate our formulation into the optimization and machine learning toolkit OMLT.•We develop an application in computer-aided molecular design of odorants.