Blood-brain barrier permeability prediction improves with combined molecular features and machine learning

Feature Space Selection and Heterogeneous Effect Estimation for Blood-Brain Barrier Permeability: A Random Forest to the Generalized Random Forest Pipeline

Machine Learning

Summary

Predicting how well drugs can cross the blood-brain barrier is important for creating medicines that work in the brain. The authors tested different ways to represent molecules and different machine learning methods to see which predicts this best. They found that combining multiple types of molecular information and using a method called Dynamic Random Forest gave the most accurate predictions. They also looked to see if some molecular traits influence permeability differently across chemical types but found no strong evidence for this after carefully accounting for confounding factors.

What this means in practice

Tested on one dataset.

Authors

Tshemollo Rapolai, Seite Makgai, Mohammad Arashi

Abstract

Predicting blood-brain barrier (BBB) permeability is critical for central nervous system drug discovery. Using the MoleculeNet BBBP dataset (n = 2039), this study systematically ablates molecular feature spaces to isolate featurisation from model architecture. We evaluate three feature families (Morgan fingerprints, RDKit physicochemical descriptors, SMILES bigrams) across four learning algorithms. Results demonstrate that predictive performance depends jointly on feature representation and algorithm. Dynamic Random Forest using combined features achieved the highest mean AUC (0.970, 95% CI: 0.963-0.977). Second, this optimal representation enables exploratory estimation of heterogeneous associations between molecular structure and BBB permeability using Generalized Random Forests. Constructing a pseudo-treatment from a LogP median split, we applied double/debiased machine learning to account for confounding. Orthogonalization substantially attenuates the heterogeneity detected by naive causal forests; no conditional effects remained significant after false discovery rate correction (smallest adjusted p = 0.082). Furthermore, orthogonalized feature importance shifted toward residual structural information in SMILES bigrams. Ultimately, once observed confounding is properly accounted for, evidence that LogP-BBB associations vary systematically across chemical space is insufficient. This underscores that feature representation and model architecture are coupled design choices, and that unorthogonalized causal forests risk overstating genuine treatment effect heterogeneity.
View PDFOpen arXiv