Linear and Nonlinear Quantitative StructureProperty Relationship Models for Solubility of Some Anthraquinone, Anthrone and Xanthone Derivatives in Supercritical Carbon Dioxide

• Published in: Analytica chimica acta Vol. 610; no. 1; p. 25
• Main Authors: Hemmateenejad, Bahram, Shamsipur, Mojtaba, Miri, Ramin, Elyasi, Maryam, oghinia, Farzaneh, Sharghi, Hashem
• Format: Journal Article
• Language: English
• Published: 01.03.2008
• ISSN: 0003-2670

Linear and nonlinear quantitative structureproperty relationship models for solubility of some anthraquinone, anthrone and xanthone derivatives in supercritical carbon dioxide were presented. The chemical structure of molecules was constructed using Hyperchem package and optimized by the AM1 semi-empirical method of the software. A total of 108 electronic, chemical, topological and constitutional descriptors were calculated for each molecule. In comparison with conventional stepwise selection method, combined data splitting feature selection (CDFS) resulted in simpler models with higher prediction ability. Both linear (i.e., MLR) and nonlinear (ANN) methods were employed to discover the structuresolubility relationships and more accurate results were found the latter. Application of the models to the external test set compounds revealed that they could be used to predict the solubility of the new compounds in the various temperatures and pressures with very low prediction error.