Alignment-free descriptors for quantitative structure-rate constant relationships of [4+2] cycloadditions

A reliable quantitative model for estimation of reactivities of Diels-Alder reactions has been developed. A linear and a non-linear method were used to predict rate constants. Computationally efficient descriptors (including chemical graph indices, intermolecular binding distances, molecular surface areas and electrostatic potentials (ESP), ESP atom value autocorrelation functions, topological (2D) or topographical parameters) with artificial neural networks as well as calculated IR-spectra (Sybyl's EVA) with partial least square regression were used to build the QSPR equations. Both sets of descriptors are invariant to roto-translation.