Chain melting temperature estimation for phosphatidyl cholines by quantum mechanically derived quantitative structure property relationships

Geometries for 62 phosphatidylcholines (PC) were optimized using the AM1 semiempirical quantum mechanical method. Results obtained from these calculations were used to calculate 463 descriptors for each molecule. Quantitative Structure Property Relationships (QSPR) were developed from these descriptors to predict chain melting temperatures (Tm) for the 41 PCs in the training set. After screening each QSPR for statistical validity, the Tm values predicted by each statistically valid QSPR were compared to corresponding Tm values extracted from the literature. The most predictive, chemically meaningful QSPR provided Tm values which agreed with literature values to within experimental error. This QSPR was used to predict Tm values for the remaining 21 PCs to provide external validation for the model. These values also agreed with literature values to within experimental error. The descriptor developed by the final QSPR was the second order average information content, a topological information-theoretical descriptor.