Figure 1.

Schematic diagram of logBB prediction model construction. (A) LogBB dataset collection and features preprocessing. (B) Model was trained on a complete dataset and further evaluated on external dataset, and average model performance was also measured by applying 10-fold cross-validation. (C) Independent test dataset was used to compare the performance of our model with other publicly available quantitative models. (D) Binarized test dataset and additional BBB-permeable drug library were used to compare the performance of our model as a qualitative model with publicly available qualitative models.