Prediction of the normal boiling point of oxygen containing organic compounds using quantitative structure–property relationship strategy

Quantitative structure–property relationship (QSPR) models were applied to predict the normal boiling point (NBP) of oxygen containing organic compounds, including alcohols, phenols, ethers, aldehydes, ketones, carboxylic acids and esters. The total 432 compounds were divided into 3 subsets according to their structure features. For each subset, 8 significant descriptors were selected from the pool of descriptors. Sequentially, the multiple linear regression (MLR) method as well as the non-linear radial basis network (RBN) was used to correlate and predict the NBP of the compounds. RBN model showed higher accuracy with respect to MLR model and Constantinou-Gani (C-G) group contribution method. Comparison with previous QSPR models indicated that the present models could be more general for NBP prediction of organic compounds with certain oxygen containing functional group. In addition, QSPR models for all the 432 compounds were also deduced, and the results confirmed that RBN model performed better in the field of QSPR modeling.