Explaining blood-brain barrier permeability of small molecules by integrated comparison of different transport mechanisms
The human blood-brain barrier (BBB) represents a major obstacle to the transfer of drugs into the central nervous system (CNS). Therefore, effective therapeutics for many CNS diseases are lacking, with only the minority of small molecules passing through the BBB. Substances may cross the BBB through different transport mechanisms includingwhich includes passive diffusion, active influx and efflux transport. Since experimental evaluation and optimization of CNS drugs is time-consuming and expensive, alternative research methods such as computational prediction models could accelerate CNS drug development. Here, we studied BBB permeability using previously published and self-curated data focusing on different aspects on BBB transport, namely influx, efflux and passive diffusion. Using these datasets, we created Prediction models, either based on physicochemical properties or molecular substructures, to define how active and passive transport mechanisms contribute to BBB permeability. Decision trees were used to determine the importance of either physicochemical features or substructures contributing to the respective transport mechanism. Molecular substructures were independently confirmed by Tanimoto similarity analysis of groups of transported compounds versus non-transported compounds. Features associated with BBB transport to the brain were subsequently compared to a benchmark model based on CNS drugs to assess their relative contribution to the sum of all transport mechanisms. Our results imply that different mechanisms apply to different subgroups of compounds in order to reach the brain. Features that predict passive diffusion overlap to the largest extend with features that explain BBB permeation or define CNS-active drugs. Furthermore, we identified a number of molecular substructures that positively or negatively predict influx and membrane diffusion, where hydrophilic beta-lactam ring presence is negatively associated with transport and where corticosteroid-derivatives and close vicinity of hydrogen bond donors/acceptors are positively associated with transport. To conclude, our results can provide significantrelevant guidance towards repurposing or discovery of new CNS therapeutics BBB permeable compounds through optimally adaptingmatching features of compounds to different BBB transport mechanisms.