Exploring blood-brain barrier passage using atomic weighted vector and machine learning.

Area of Science:

• Computational chemistry and cheminformatics
• Pharmacology and drug discovery
• Artificial intelligence in medicine

Background:

• Accurate prediction of blood-brain barrier (BBB) permeation is crucial for developing effective central nervous system (CNS) drugs.
• Leveraging molecular properties through computational tools can enhance drug development pipelines.
• Existing methods for predicting BBB passage require optimization for improved accuracy and efficiency.

Purpose of the Study:

• To investigate the predictive power of molecular properties, derived from MD-LOVIs software, for estimating compound BBB permeability.
• To apply and evaluate various machine learning (ML) models, including classification and regression techniques, for predicting BBB passage and molecular activity.
• To identify the most effective ML models for accurately forecasting a compound's ability to cross the BBB.

Main Methods:

• Utilized MD-LOVIs software to generate molecular descriptors from compound structures.
• Employed a range of machine learning algorithms: Gradient Boosting Machines (GBM), Generalized Linear Models (GLM), Support Vector Machines (SVM) with polynomial kernels, Random Forests (RF), ensemble regression models, and instance-based learning algorithms.
• Trained and validated ML models on diverse datasets, reporting performance metrics such as accuracy and R-squared values.

Main Results:

• Classification models demonstrated high accuracy in predicting BBB passage, with SVMPoly variants achieving up to 0.980 accuracy.
• Regression models showed strong performance in predicting molecular activity, with ES-RLM-AG achieving an R-squared value of 0.902.
• Specific models like GBM-AWV, GLM-CN, SVMPoly variants, ES-RLM-AG, and IB-MLP proved highly effective, confirming the utility of ML in this domain.

Conclusions:

• Machine learning techniques, combined with molecular property analysis, offer a powerful approach for predicting blood-brain barrier permeability.
• The study highlights the effectiveness of specific classification and regression models in forecasting CNS drug potential.
• These findings can significantly accelerate the early stages of drug discovery for neurological disorders by improving the selection of viable drug candidates.