Relative recovery (Re) is one of the major concerns of microdialysis, a valuable sampling technique which can continuously collect unbound drugs in blood and most tissues. For a given microdialysis probe, the recovery of every compound from the probe is related to its structural characteristics and physicochemical property if the experiment condition is fixed. In this work, quantitative structure-property relationship (QSPR) models using multiple linear regression (MLR) and support vector machine (SVM) methods were setup to excavate the relationships of microdialysis Re of compounds and their molecular descriptors which capture the structural characteristics of molecules for a series of flavone derivatives. As result, significant statistical parameters (MLR model: R² = 0.9268 (correlation coefficient), Q²LOO = 0.8572 (explained variance in prediction) and Q²ext = 0.8639 (external explained variance), and SVM model: R² = 0.9383 and Q²ext = 0.8536) were obtained, indicating good stability and predictive ability of the models. Therefore, it seems feasible to predict the microdialysis relative recovery of some compounds from their molecular descriptors. This investigation was an innovative trial and can provide new methods for researching the microdialysis recovery of the compounds.
quantitative structure-property relationship (QSPR); microdialysis; relative recovery; multiple linear regression; support vector machine; flavone derivatives
