Physicochemical determinants of blood brain barrier penetrating molecules

The blood-brain barrier (BBB) is an essential physiological barrier that regulates the transport of substances from the circulation to the brain. Accurate prediction of BBB permeability is essential for understanding drug delivery to the brain and for developing effective therapies for neurological disorders.Clinical experiments have provided a more accurate measure of BBB permeability.Nevertheless, these methods take time and are labor-intensive.Consequently, several computational methods have attempted to predict BBB permeability; however, their accuracy remains a challenge.Within the scope of this investigation, we provide a novel strategy for enhancing the precision of BBB permeability prediction models. Our model integrates a diverse set of molecular descriptors and employs advanced machine-learning algorithms to identify complex connections between chemical compounds and BBB permeability.By using a large dataset of experimental observations and various resampling techniques, we increased the prediction performance of our model. Different machine learning algorithms (Random Forest (RF) and Gradient Boosting Machine (GBM)) algorithms were used and further analyzed using model agnostic interpretation methods, to accurately predict BBB permeability. The highest accuracy of 92.5% was obtained by RF with feature set of JOELib descriptor (SMOTE oversampled), followed by RF with feature set of JOELib descriptor (GAN oversampled) and the accuracy of 92.1%.Shapley plots, ALE plots, and variable importance plots (VIP) were used to depict the significance of the features.