Development of QSPR Method for Prediction of Physicochemical Properties of Molecules

초록

In the design of new materials, it is necessary to examine their physicochemical properties before synthesis. Heats of formation and solid density are two basic quantities that can be used to estimate the energy contents and crystal packing. Numerous studies have been done to predict these properties and have obtained some success using quantitative structure property relationship (QSPR) approaches. In this work, we developed a new approach that can be employed any physicochemical properties of molecules. In usual QSPR studies, it is customary to calculate various molecular descriptors from the molecular formula and three-dimensional structure. This work is done for every molecule in the training set. Then a linear or non-linear object function is derived using statistical approach. In most cases, a linear correlation is tried and the goodness of the fit is judged from the regression coefficient and root-mean square deviation (RMSD) between experimental and predicted values.