In Silico Models to Predict Brain Uptake

Abstract

A rapid and high-throughput assessment of brain penetration is important in the ‘hits to lead’ and ‘lead optimization’ phases of the drug discovery process. It is important not only for identifying compounds that penetrate the blood–brain barrier in sufficiently high concentration to exhibit a therapeutic effect, but also in the design of compounds with minimal brain penetration, which is important for reducing unwanted side effects such as sedation. In silico models are important tools in assessing brain penetration as they have the potential to make predictions on large numbers of compounds. In addition, the ability to make predictions in terms of brain penetration for virtual compounds allows in silico methods to be used in the design of compounds. For this, the use of simple, interpretable, and quick to calculate properties is important. Polar surface area and hydrogen bonding counts are examples of properties frequently utilized in this context. There are a variety of in vivo measures of brain penetration that are used in drug discovery to assess compounds and many in silico models have been built on these endpoints. Brain perfusion techniques give a measure of the rate of entry of compound into the brain. Although the perfusion data sets are small, results show that perfusion rate correlates poorly with log Poct but exhibits a negative correlation with hydrogen bonding parameters. The effect of compound ionization on perfusion currently appears to be complex. The most commonly used measure of brain penetration is the ratio of compound in the brain to compound in the blood or plasma (BB) and this is the measure that most literature models have been built from. There are consistent properties that emerge from these models as either positively or negatively influencing the brain penetration of compounds. Descriptors encoding molecular size correlate positively with brain penetration whereas descriptors describing the polar nature of compounds (polar surface area and hydrogen bond acceptor and donor groups) correlate negatively with brain penetration. The percentage of compound ionized at physiological pH does not appear to be an influencing factor in brain penetration. It is, however, observed that acids have lower brain penetration than would be predicted on the basis of their other molecular properties. Owing to the limited experimental in vivo data available many workers have used knowledge of whether compounds have CNS activity or CNS side effects (CNS+) compared with compounds that have no CNS activity or side effects (CNS−) to determine factors affecting brain penetration. These analyses show the same trends between molecular descriptors of polarity and size with brain penetration (CNS+/−) as is observed for measured BB values. The rules of thumb derived from these analyses provide a useful way of ‘screening’ or ‘filtering’ large compound sets in the early stages of drug discovery. The limitation of this approach is in the assessment of nonbrain penetrating compounds and this difficulty should not be underestimated.