A series of Aminopyrimidine-Based derivatives was under principle component analysis, linear regression, nonlinear regression, and partial least squares analysis to establish a QSAR model of CXCR4 antagonists. We proposed three models, statistically investigated, and properly validated internally and externally. The best model is the PLS model, for it demonstrated the highest prediction ability, the lowest mean squared error, and an accepted coefficient of determination (R2=0.735; R2test=0.774;

more »

... MSE=0.127); Tropsha and Golbraikh's parameters were calculated and found within the acceptance threshold. The Y-randomization method was employed to verify the absence of any chance correlation. As for the visualization of the applicability domain, we used William's plot, where we found no outliers. All of the obtained results confirmed the prediction reliability of the chosen model.