Application of PAMPA technique and in silico blood-brain barrier permeation prediction of protein kinase inhibitors and selected CNS drugs

Objective The study and prediction of blood-brain barrier (BBB) permeability is an important part of drug discovery, both for drugs that target central nervous system (CNS) disorders and those that do not [1]. One of the most widely used experimental methods for permeation determination is the Parallel Artificial Membrane Permeability Assay (PAMPA). After calculating the effective permeability from PAMPA, various quantitative structure-property relationships (QSPR) can be developed using in silico calculated descriptors of compounds. In this study, we developed and analyzed several QSPR models for two groups of compounds – a group of protein kinase (PK) inhibitors and a set of CNS-active compounds. Methods The group of PK inhibitors consisted of 17 commercially available substances and 17 experimental compounds [2], while the set of CNS drugs comprised 12 commercially available and 6 experimental compounds [3]. The effective permeabilities of all compounds were determined using PAMPA followed by high-performance liquid chromatography (HPLC). For the set of PK inhibitors, multiple linear regression (MLR), support vector machine regression (SVM) and artificial neural networks (ANN) were used to develop QSPR models with effective permeability as the dependent variable, while for the set of CNS drugs, MLR, SVM and partial least squares (PLS) regression methods were used and all of the developed models were validated. Results According to the effective permeabilities determined with PAMPA, the group of PK inhibitors contained a variety of low, medium and high permeability compounds. As expected, almost all CNS drugs were classified as highly permeable. For the set of PK inhibitors, all 3 mathematical methods yielded QSPR models with comparable performances, with the SVM model meeting all critical validation parameters. For the set of CNS drugs, PLS and SVM methods provided optimal QSPR models, with the SVM model being the better one. Conclusions In this study, the PAMPA-BBB technique was used to determine the ...