Improved Prediction of Blood-Brain Barrier Permeability Through Machine Learning with Combined Use of Molecular Property-Based Descriptors and Fingerprints

Abstract

Blood-brain barrier (BBB) permeability of a compound determines whether the compound can effectively enter the brain. It is an essential property which must be accounted for in drug discovery with a target in the brain. Several computational methods have been used to predict the BBB permeability. In particular, support vector machine (SVM), which is a kernel-based machine learning method, has been used popularly in this field. For SVM training and prediction, the compounds are characterized by molecular descriptors. Some SVM models were based on the use of molecular property-based descriptors (including 1D, 2D, and 3D descriptors) or fragment-based descriptors (known as the fingerprints of a molecule). The selection of descriptors is critical for the performance of a SVM model. In this study, we aimed to develop a generally applicable new SVM model by combining all of the features of the molecular property-based descriptors and fingerprints to improve the accuracy for the BBB permeability prediction. The results indicate that our SVM model has improved accuracy compared to the currently available models of the BBB permeability prediction.

Keywords: blood–brain barrier permeability; fingerprint; modeling; molecular descriptor; physical property.

Figure 1.

Work flowchart of the SVM training and validating.

Figure 2.

Comparison of the quality of the SVM models with different descriptor vectors in the prediction for the test set. (A) Training set included 80% compounds selected from dataset A by the KS method, and the test set included the remaining 20% compounds from dataset A. (B) Training set included 80% compounds selected from dataset A by the RS method, and the test set included the remaining 20% compounds from dataset A. The average prediction accuracy and standard deviation were caclulated from 5 independent training sets generated by the RS method (containing 80% compounds of the whole dataset). (C) Training set and test sets were selected from dataset B with the KS method. (D) The training set and test sets were selected from dataset B with the RS method.