QSAR modeling of P-glycoprotein (Pgp) recognition as a new strategy for the development of new anticonvulsant agents for the treatment of refractory epilepsy

ALAN TALEVI; MAURICIO E. DI IANNI; LUIS E. BRUNO-BLANCH

Honolulu

Congreso; Pacifichem 2010; 2010

American Chemical Society

Epilepsy is the most common chronic disorder of the central nervous system (CNS). According to estimations from WHO, 90% of the epileptic population comes from developing countries and 30% of the patients do not achieve control of the symptoms with available antiepileptic durgs, condition known as refractory epilepsy [1]. Recent reports reveal an association between refractory epilepsy and over-expression of efflux transporter glycoprotein P at the blood-brain barrier [2]. Therefore, development of new Pgp-nonsubstrates antiepileptic drugs appears as an interesting alternative for the treatment of refractory epilepsy. From a 125-drug training set composed by Pgp-nonsubstrates and substrates, a QSAR model containing four topological descriptors was derived through linear discriminant analysis and validated through standard validation procedures. The global percentage of good classifications in the training set was 78.4%. The average global percentage of good classifications in leave group out (LGO) cross-validation was 76.6% (sd = 1.8). The average global percentage of good classifications for the randomizations models is 60.8% (sd = 2.7). As expected, this latter approaches random classification. The global percentage of good classifications in a 125-compound test set was 77.6% (similar to that on the training set, indicating low chance of overfitting). The results suggest the model is robust and has good generalizability. An almost 80% of good classifications represents a good result if one takes into account that Pgp is characterized by broad substrate specificity and thus common features among its substrates are not easy to identify. The model might be used as an additional filter in virtual screening campaigns to identify new drug candidates without affinity for Pgp.