摘要：A simplified in silico QSAR study, involving more than 50 molecules that are commonly used as UV filters in sunscreens, is presented. Details of methodology are described and illustrated step by step and the use of available tools is demonstrated, including data set selection, generation of molecular structure files, selection of computational methods, assessment of the relative importance of molecular properties, generation of theoretical UV-Vis spectra and construction of a QSAR model. With the aim of introducing the QSAR methodology at undergraduate level, some simplifications have been made. Multiple Linear Regression (MLR) has been selected as a modeling strategy to correlate the molecular descriptors and the sunlight blocking effect. The full area under the UV-Vis spectrum has been used as a simple measure of the protection factor. Theoretical UV-Vis spectra have been generated by using the configuration interaction (CI) approximation. The ground state molecular geometry has been optimized by the semi-empirical ZINDO computational method. Inclusion of a solvent model in the optimization of molecular geometry to account for solvatochromic effects has not been considered. On the other hand, total number of rings, number of aromatic rings, molecular dipole moment and HOMO/LUMO energy gap, are considered to be significant parameters in the MRL model. The HOMO/LUMO energy gap is the variable of choice when optimized single linear regression analysis is carried out separately for each of the functional groups of the compounds. The great variety of tools and procedures used in building QSAR models has been found to provide ample scope for active involvement by the students, exposing them to a fairly broad array of modern methods in areas like drug discovery and toxicology.
