A machine learning based computer-aided molecular design/screening methodology for fragrance molecules. (12th July 2018)

• Record Type: Journal Article
• Authors: Zhang, Lei; Mao, Haitao; Liu, Linlin; Du, Jian; Gani, Rafiqul
• Is Part Of: Computers & chemical engineering. Volume 115(2018)
• Page Start: 295
• Page End: 308
• Publication Date: 2018-07-12
• Subjects: Computer-aided molecular design -- Fragrance -- Machine learning -- Group contribution method -- Product property; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2018.04.018
• Languages: English
• ISSNs: 0098-1354

Abstract

Hightlights: A computer-aided molecular design/screening method is developed for fragrance molecules. The odor of molecules are predicted using a data driven machine learning approach. A MILP/MINLP model is established and decomposition-based solution approach is applied for the design/screening. Case studies highlighting the effectiveness of the developed model are presented. Abstract: Although the business of flavors and fragrances has become a multibillion dollar market, the design/screening of fragrances still relies on the experience of specialists as well as available odor databases. Potentially better products, however, could be missed when employing this approach. Therefore, a computer-aided molecular design/screening method is developed in this work for the design and screening of fragrance molecules as an important first step. In this method, the odor of the molecules are predicted using a data driven machine learning approach, while a group contribution based method is employed for prediction of important physical properties, such as, vapor pressure, solubility parameter and viscosity. A MILP/MINLP model is established for the design and screening of fragrance molecules. Decomposition-based solution approach is used to obtain the optimal result. Finally, case studies are presented to highlight the application of the proposed fragrance design/screening method.