An In Silico Model for Classification of P-Glycoprotein Substrates and Non-Substrates

Abstract: P-glycoprotein (P-gp) is one of the most important enzymes in ATP transporter family. It transports various foreign substances actively out of cells and, in this way, greatly influencing the ADME (absorption, distribution, metabolism and excretion) properties of drugs. P-gp has a very wide spectrum of substrates, whose structures are heavily diversified. In the present work, we attempted to classify more than 150 compounds which have been experimentally identified as P-gp substrates or non-substrates. By using discriminant analysis, an in silico model was obtained with satisfactory statistical results. For both the original grouped cases and the cross-validated grouped cases more than 80% cases were correctly classified. When being validated on the test set, it also exhibited good predictability. Our model not only proved the applicability of discriminant analysis for classification of P-gp substrates and non-substrates, but also was helpful for further screening of P-gp receptors