Objective:

In this study, we utilize various machine learning, deep learning, and transfer learning methods to predict the blood-brain barrier (BBB) permeability of public drug datasets. 

Background:

The BBB is a semi-permeable boundary in the central nervous system (CNS) that maintains homeostasis of the brain microenvironment by selection of molecules that are allowed to pass through. The limitation is that clinical experiments, the current standard of determining which of these compounds effectively cross, are time consuming and labor intensive. Machine learning and other predictive methods offer the ability to accelerate this process by building models trained on datasets with known permeability. 

Design/Methods:

The 7,807 compounds used in this study are from one of the largest public BBB datasets with categorical values. We trained and validated models on this dataset with methods including support vector machines (SVMs), deep neural networks (DNNs), and graph convolutional neural networks (GCNNs). The transfer learning model consisted of an initial DNN trained to a single quantum chemical property that was then subsequently fine-tuned to the task of BBB permeability after the appendage of additional neural network layers. 

Results:

The prediction accuracies on the validation set for SVM, DNN, GCNN, transfer learning of polarizability, and transfer learning of dipole moment were 82.33%, 83.09%, 87.14%, 76.89%, and 70.23%, respectively. 

Conclusions:

Currently, GCCN’s demonstrate the best predictive capability. This underscores the importance of this algorithm's message passing phase in learning the molecular properties most contributory to the predictive task. Future work consists of exploring the correlation between quantum chemical properties and BBB permeability to aid in the success of transfer learning model development. This study demonstrates the ability of machine learning to offer insight into the CNS-activity of certain drug molecules and identify key molecular features that may guide the design of future therapeutics.