Quantitative Structure-Activity Relationship and Classification Analysis of Diaryl Ureas Against Vascular Endothelial Growth Factor Receptor-2 Kinase Using Linear and Non-Linear Models

摘要: Quantitative structure-activity relationship analysis has been carried out for 74 diaryl ureas including aminobenzoisoxazole ureas, aminoindazole aminopyrazolopyridine against vascular endothelial growth factor receptor-2 kinase using both linear and non-linear models. Considering simplicity predictivity, multivariate regression was first employed in combination with various variable selection methods, forward selection, genetic algorithm enhanced replacement method based on descriptors generated by e-dragon software. Another model support vector also constructed compared. Performances of these models are rigorously validated leave-one-out cross-validation, fivefold cross-validation external validation. The significantly outperforms the others R(2) = 0.813 R(2)(pred) 0.809. Robustness predictive ability this is prudently evaluated. Moreover, to find most significant features associated difference between highly active compounds moderate ones, two classification discriminant machine were further developed. performance analysis, validation prediction accuracy reaching 0.838 0.857, respectively. resulting could act as an efficient strategy estimating inhibiting activity novel provide some insights into structural related biological compounds.