Integrating Safety and Health Indexes in Computer-Aided Molecular Design Under Property Prediction Uncertainty
Computer-Aided Molecular Design (CAMD) is a powerful tool for designing promising molecules with the targeted properties. Besides properties, the safety and health aspects must also be considered during design stage to ensure the synthesized molecules do not cause much harm. The molecules are assessed by several safety and health parameters, which are measured using physicochemical properties that can be estimated via property prediction methods. Each parameter/sub-index is allocated with a penalty score based on the property to represent the hazard level of molecules. However, uncertainty in the prediction models can adversely affect the accuracy of scores. The main highlight of this paper is to manage the effect of uncertainty on the safety and health sub-indexes, by improving the scores at the uncertain region for enhanced measurement of molecular hazards. A case study on solvent design for carotenoids extraction from Palm Pressed Fiber (PPF) has been considered with the proposed methodology.