Systematic Framework for Molecular Design: Methodology and Applications.

摘要

We propose a new framework to automate, augment, and accelerate steps in computer-aided molecular design (CAMD). We also extend the scope of CAMD methods beyond group contribution models by employing accurate and extensive property models with a wide application domain. The molecular design problem is tackled in three stages: 1) composition design, 2) structure determination, and 3) extended design. Composition identification and structure determination are decoupled to achieve computational efficiency. Using approximate group contribution methods in the first stage, molecular compositions that fit the relaxed design targets are identified. In the second stage, isomer structures of solution compositions are determined systematically and structure-based property corrections are utilized to refine the solution pool. In the final stage, the design is extended beyond the scope of group contribution methods by utilizing problem-specific property models. As we progress through these stages, the information available about candidates increases while the size of the solution pool decreases. At each stage, optimization algorithms generate a large and diverse pool of candidate designs using an assortment of property models. The molecule generation machinery and external property models are bundled together to create an easy-to-use software.;The use of integer optimization techniques instead of enumeration at each stage, leads to an efficient algorithm, unrestricted in terms of the number of solutions sought and the maximum size of the molecule sought. By exploiting linearity through suitable problem formulations, large solution sets are easily handled. Unique solutions are generated effectively with cuts to handle redundancy in chemical space. These optimization-based models lead to orders of magnitude decrease in solution time over enumeration or MINLP approaches. Along with the group-contribution model used to build the molecules and predict their basic properties, higher-order structural descriptors and missing group contributions are also included to achieve high accuracy of property prediction. The modular nature of our approach also allows use of an assortment of property models for a specific application. Many design criteria that could not be included in simple molecular design can be incorporated. The proposed algorithm can be used as a stand-alone tool or it can be expanded as a part of more complex design problems like mixture design.;The developed solution techniques are applied to a variety of industrial problems. We design solvents for pharmaceutical drugs and solvents for degreasing process. The results for these case studies match the previously reported solvent families. The problem of designing automobile refrigerants is also solved to demonstrate the effectiveness of the framework. The identification of state-of-the-art refrigerants validates our approach. Many replacement molecules are designed for use as secondary heat transfer fluids in supermarkets. These arc compared on the basis of their heat transfer performance and are being analyzed for their global warming potential. We also identify existing compounds as replacement refrigerants for the growing market of electronic cooling systems. We design synthetic base fluids for deep drilling and investigate them for toxicity. The results not only match current base fluids but also provide better alternatives with improved economic features. We also present preliminary work on designing quaternary ammonium surfactants for oil-based and water-based drilling fluids.