Taking drug-loaded micelles self-assembled from the amphiphilic polymers as a case

the quantitative structure-property relationship (QSPR) of chemical products was studied on the basis of the cross-scale method from macroscopic to microscopic in this work. A block unit autocorrelation method to calculate the descriptors for characterization of amphiphilic polymer structure was proposed. A series of QSPR models to predict the drug-loading capacity of polymeric micelles at different drug/polymer ratios were established by combining genetic function approximation (GFA) algorithm and multiple linear regression (MLR). The reliability

stability and predictive ability of models were evaluated via internal validation (R2>0.854

Qloo-cv>0.651

RMSE<0.089)

Y-randomization test

external validation (Qext2>0.629) as well as application domain definition analysis. The effect of blocks and topological structure of polymers on drug-loading capacity was well explained by descriptors and drug-loaded mechanism analysis. The QSPR models were hopefully used to provide a guide for new polymer design and drug-loading capacity prediction. Also this cross-scale method would be expected to promote the design and development of novel complex structured products.