DLBBBP: blood-brain barrier permeability prediction based on molecular property using deep learning

The ability of compounds to pass through the blood-brain barrier is an important factor in drug development related to the central nervous system. Therefore, predicting the blood-brain barrier permeability of compounds at high throughput and providing appropriate candidate compounds are crucial for the development of related drugs. Although traditional experimental methods can also predict the blood-brain barrier permeability of compounds, they are costly and have long time cycle. To assist in related research, this article proposes a neural network model using deep learning algorithm to complete the task of predicting blood-brain barrier permeability of compounds, and names it DLBBBP. In DLBBBP, various compounds are characterized using molecular graphs and MACCS molecular fingerprints. Specifically, we conducted feature complementarity processing on MACCS, removed information about molecular substructures to prevent duplication and omission. By extracting features from the MACCS molecular fingerprints and molecular graphs of the compounds, we predict the blood-brain barrier permeability of the compounds and compare the results with some current deep learning models and machine learning methods. The validation results verify that the model's performance is better than state of the art models, and the prediction of the blood-brain barrier permeability of the compounds is accurate and effective. Therefore, it is believed that this model has great potential in the field of drug development.