Quantitative Modeling for Prediction of Critical Temperature of Refrigerant Compounds

The quantitative structure-property relationship (QSPR) method is used to develop the correlation between structures of refrigerants (198 compounds) and their critical temperature. Molecular descriptors calculated from structure alone were used to represent molecular structures. A subset of the calculated descriptors selected using a genetic algorithm (GA) was used in the QSPR model development. Multiple linear regressions (MLR) were utilized to construct the linear prediction model. The prediction result agrees well with the experimental value of this property. The comparison results indicate the superiority of the presented model and reveal that it can be effectively used to predict the critical temperatures of refrigerant compounds from the molecular structures alone. The stability and predictivity of the proposed model were validated using internal validation, external validation and Y-scrambling. Application of the developed model to a testing set of 39 organic compounds demonstrates that the new model is reliable with good predictive accuracy and simple formulation. The R2, RMSEtr and Q2loo values for the training set were 0.9752, 13.8994 and 0.9742; Q2ext and RMSEpr for test set were 0.9766 and 12.8654 for GA-MLR model, respectively. The prediction results are in good agreement with the experimental values. In addition, the applicability domain (AD) of the model was analyzed based on the Williams plot.