Building predictive unbound brain-to-plasma concentration ratio (Kp,uu,brain) models

Abstract: Abstract The blood-brain barrier (BBB) constitutes a dynamic membrane primarily evolved to protect the brain from exposure to harmful xenobiotics. The distribution of synthesized drugs across the blood-brain barrier (BBB) is a vital parameter to consider in drug discovery projects involving a central nervous system (CNS) target, since the molecules should be capable of crossing the major hurdle, BBB. In contrast, the peripherally acting drugs have to be designed optimally to minimize brain exposure which could possibly result in undue side effects. It is thus important to establish the BBB permeability of molecules early in the drug discovery pipeline. Previously, most of the in-silico attempts for the prediction of brain exposure have relied on the total drug distribution between the blood plasma and the brain. However, it is now understood that the unbound brain-to-plasma concentration ratio ( Kp,uu,brain) is the parameter that precisely indicates the BBB availability of compounds. Kp,uu,brain describes the free drug concentration of the drug molecule in the brain, which, according to the free drug hypothesis, is the parameter that causes the relevant pharmacological response at the target site. Current work involves revisiting a model built in 2011 and uploaded in an in-house server and checking for its performance on the data collected since then. This gave a satisfying result showing the stability of the model. The old dataset was then further extended with the temporal dataset in order to update the model. This is important to maintain a substantial chemical space so as to ensure a good predictability with unknown data. Using other methods and descriptors not used in the previous study, a further improvement in the model performance was achieved. Attempts were also made in order to interpret the model by identifying the most influential descriptors in the model.