Prediction of the Blood–Brain Barrier (BBB) Permeability of Chemicals Based on Machine-Learning and Ensemble Methods

• DOI: 10.1021/acs.chemrestox.0c00343
• Publication Date: 2021-05-28T12:49:44Z
• Authors: Lili Liu; Li Zhang; Huawei Feng; Shimeng Li; Miao Liu; Jian Zhao; Hongsheng Liu
• Keywords: Ensemble Learning; Ensemble forecasting; Molecular descriptor; Predictive modelling

The ability of chemicals to enter the blood–brain barrier (BBB) is a key factor for central nervous system (CNS) drug development. Although many models BBB permeability prediction have been developed, they insufficient accuracy (ACC) and sensitivity (SEN). To improve performance, ensemble were built predict compounds. In this study, in silico ensemble-learning developed using 3 machine-learning algorithms 9 molecular fingerprints from 1757 (integrated 2 published data sets) permeability. best performance base classifier was achieved by model based on an random forest (RF) MACCS fingerprint with ACC 0.910, area under receiver-operating characteristic (ROC) curve (AUC) 0.957, SEN 0.927, specificity 0.867 5-fold cross-validation. better than that most classifiers. final has also demonstrated good external validation can be used early screening CNS drugs.