Atom-type-based AI topological indices: application in modeling boiling points of aliphatic esters

In this work, modeling normal boiling points (NBPs) of a group of aliphatic esters using a combination of the modified Xu (mXu) and the atom-type-based AI topological indices was studied. At first, simple linear regression model was developed using mXu index alone and the statistics of squared correlation coefficient, Fisher ratio and standard error were 19..0, 0.88891 and 79.5, respectively. Significant decrease in the standard error of the model (about .12) was observed upon introduction of the topological indices of AI(–CH8) and AI(–O–). Statistical validity and stability of the three-parameter regression model was tested using external validation technique by dividing the entire data set into five subsets and predicting NBPs for each subset from the other four as training sets. Average standard error of 8978 for calibration set and 8980 for prediction set revealed high potential of the topological indices for developing reliable model for prediction of NBPs of aliphatic esters. Based on the results, fraction contribution of the topological indices entered the model decreased in the order of mXu> AI(–CH8)> AI(–O–) showing that NBPs of aliphatic esters are mainly dominated by molecular size, and degree of branching and polar character of the molecules have smaller contributions to the normal boiling points.

Keywords:

Quantitative Structure–Property Relationship , Modeling , Atom-Type-Based AI Topological Indices , Normal Boiling Point