In silico model to assist drug design for the treatment of refractory diseases

MELISA E. GANTNER; LUCAS N. ALBERCA; ANDREW G. MERCADER; LUIS E. BRUNO-BLANCH; ALAN TALEVI

Ciudad Autónoma de Buenos Aires

Workshop; Biowaivers 2015 Implementación de bioexenciones basadas en el Sistema de Clasificación Biofarmacéutica; 2015

International Pharmaceutical Federation (FIP), Cátedra de Control de Calidad de Medicamentos-FCE-UNLP, Cátedra de Toxicología Farmacéutica-FCE-UNLP

Objective: Refractory diseases have been associated with an overexpression of members of the ABC (ATP-binding cassette) transporters superfamily such as P-glycoprotein (Pgp) and Breast Cancer Resistance Protein (BCRP). A number of studies demonstrate that BCRP is the most abundantly expressed ABC efflux transporter throughout the intestine and the blood-brain barrier of healthy tissues, limiting oral absorption and brain bioavailability of its substrates. Early recognition of BCRP substrates is thus essential to design novel therapeutics for the treatment of refractory diseases like cancer, HIV, epilepsy and other central nervous system conditions linked to BCRP-mediated multidrug resistance issues. Therefore, the aim of this work is the development of an in silico model based on conformation-independent molecular descriptors capable of discriminating between BCRP substrates and non-substrates. Materials and Methods: A data set of 262 substrates and non-substrates of BCRP was compiled from literature. Representative and balanced training and test sets were obtained through a two-step clustering process. The Enhanced Replacement Method (ERM) and the ensemble learning approaches were applied to obtain combinations of 2D linear classifiers. For the external validation, we built two different databases larger than the original test set, in order to estimate in a more realistic way the utility of our model in a real virtual screening setting. We applied Receiving Operating Characteristic (ROC) curves analysis to assess and compare the performance of the models. Results: By the application of the ERM, we generated 459 models. The combination of the 10 best individual models improves the performance compared to the best individual model. According to the computational experiments conducted over all databases, the best model ensemble has the best capacity to discriminate between BCRP substrates and non-substrates. Conclusions: The in silico model ensemble developed is a potentially valuable tool to assist the discovery of new drugs for the treatment of refractory diseases.