HIGH-PERFORMANCE LIQUID CHROMATOGRAPHY OF CHALCONES: QUANTITATIVE STRUCTURE-RETENTION RERLATIONSHIPS USING PARTIAL LEAST-SQUARES MODELING

MONTAÑA M.P.; PAPPANO N. B.; DEBATTISTA N. B.; LUCO J.M.

Viña del Mar, Chile

Congreso; 5th LATINAMERICAN CONFERENCE ON PHYSICAL ORGANIC CHEMISTRY; 1999

Chalcones have been reported to display a wide variety of biological activities. For example, several derivaties have shown antibacterial, antiviral, hepato protectant, antimutagenic,  gastric protectant, anti-inflammatory, antileishmanial and anticancer activities. It is generally accepted that fortuious or artifactual QSPR/QSAR models rnay be obtained when there exists multicollinearity amoung predictor variables and the models are derived by means of multiple linear regression (MLR) techniques. The disadvantages of the MLR method has recently been overcome through the development of a partial least squares (PLS) method which circumvents the problems of collinearity amoung the variables and also offers the advantage of handling data sets where the number of independent variables is greater than the number of observations. In this study, the multivariate Partial Least Squares Projections to Latent Structures (PLS) technique was used for rnodeling the Reverse-Phase High-Perforrnance Liquid Chromatography (RP-HPLC) retention data of 17 chalcones, which were determined with methanol-water mobile phases of different compositions. The PLS model was based on molecular descriptors which can be calculated for any compound utilizing only the knowledge of its molecular structure. The PLS analysis resulted in a model with the following statistics: r = 0.976, Q = 0.933, s = 0.076 and F = 43.63. The adequacy of the developed model was assessed by means of crossvalidation and also, by PLS modeling of the retention data of several chalcones reported by Walczak et al., which were obtained using stationary phases of different polarity (-NH2, DIOL, CN, ODS, C8). The structural interpretation of the developed PLS model was accomplished by means of comparative correlations between the nonempirical descriptors used in the model and the solvation parameters developed by Abraham. The results obained in this work provides evidence for the great potential of topological approach for developing Quantitative Structure-Retention Relationship (QSRR) models.