In Silico Prediction of CNS Bioavailability

The development of new treatments for conditions of the central nervous systems is particularly challenging, with higher attrition rates during development than in other therapeutic areas. The blood–brain barrier represents one of the major obstacles to efficient drug delivery to the brain. The pharmaceutical industry has long recognized that drug development should be approached as a multi-parameter optimization task. An effective and safe drug must balance several pharmaceutically relevant properties, including potency against its intended target(s), appropriate pharmacokinetic and safety profiles, chemical stability, and novelty. The optimal values of all these properties, when possible, are simultaneously pursued during the hit identification and hit-to-lead stages, as the pharmaceutical sector has embraced the “to fail early is to fail cheap” paradigm. Early assessment of the absorption, distribution, metabolism, and excretion (ADME) properties of a drug is performed by means of in silico filters and in vitro assays. Here, we will critically review past and present approaches to predict central nervous system bioavailability for small molecules, from hard-logic filters to multiparameter optimization, from quantitative structure-property relationships (QSPR) and machine learning to structure-based approaches that may be used to examine the interactions between small molecules and relevant blood-brain barrier transporters.