New electrotopological descriptor for prediction of boiling points of alkanes and aliphatic alcohols through artificial neural network and multiple linear regression analysis

A new molecular descriptor -Sigma q(c),, the sum of partial atomic charges on the carbon atoms of the molecule which is an electro-topological parameter is used for the prediction of boiling points of acyclic alkane isomers from C1 to C9, normal alkanes from C1 to C100 and 50 acyclic mono alcohols. The evaluated nonlinear equations correlated better than the linear equations. The electrotopological descriptor along with some graph topological molecular fragments have been evaluated for the prediction and correlation of boiling points of alkanes and alcohols using multilinear regression analysis and artificial neural network. The electrotopological parameter gave a very good correlation for the prediction of boiling points of alcohols along with one of the graph topological molecular fragments. The artificial neural network also shows a good fitting capability and a model network is evaluated. (C) 1998 Published by Elsevier Science Ltd. All rights reserved.