QSAR analysis of the effects of OATP1B1 transporter by structurally diverse natural products using a particle swarm optimization-combined multiple linear regression approach

• Natural product-drug interactions are for the first time studied.
• PSO-MLR is used for selecting optimal variables and building models.
• Good correlation between natural products and inhibitory effect was constructed.
• Several validation techniques are used to demonstrate the reliability.

The inhibitory effect of 100 structurally diverse natural products on the uptake of estrone-3-sulfate (E3S) by OATP1B1 was evaluated using a rapid transport assay. For each natural product, a total of 252 two-dimensional descriptors were calculated using ChemoPy software package developed by our group. Using a particle swarm optimization-based multiple linear regression (PSO-MLR) method, the desired descriptors were automatically selected to maximize the predictability of the IC50 values. The logarithm of IC50 values of the natural products tested ranged from 2.009 to 2.882. Based on the PSO-MLR method, nine ChemoPy descriptors were found to be important and effective for QSAR modeling. Interestingly, these descriptors suggested that the molecular complexity and electrotopological state would be important factors in determining natural products-OATP1B1 interactions. In the leave-one-out prediction, the correlation coefficient of prediction (Q) and the standard error of prediction (s) were 0.824 and 0.098, respectively. Even in an external validation, the predictions were in good agreement with experimental values (Rpred = 0.820, s = 0.109, n = 20). The proposed model, in which ChemoPy descriptors were used as molecular descriptors, was able to predict drug-OATP1B1 interactions with reasonable accuracy.