Artificial neural network approach for predicting blood brain barrier permeability based on a group contribution...

Area of Science:

• Computational chemistry
• Pharmacokinetics
• Drug discovery

Background:

• Predicting blood-brain barrier (BBB) permeability is crucial for developing effective central nervous system drugs.
• Quantitative structure-activity relationship (QSAR) models offer a computational approach to estimate BBB penetration.
• Artificial neural networks (ANNs) show promise in developing accurate predictive models.

Purpose of the Study:

• To develop a quantitative structure-activity relationship (QSAR) model for predicting blood-brain barrier (BBB) permeability.
• To utilize artificial neural networks (ANNs) in conjunction with molecular structure and property descriptors.
• To establish a reliable computational tool for assessing drug candidates' ability to cross the BBB.

Main Methods:

• A database of 300 compounds was compiled.
• Fifty-two structure descriptors derived from the UNIFAC group contribution method and eight molecular property descriptors were employed as input features.
• Artificial neural networks (ANNs) were trained using these descriptors to predict logBB values.

Main Results:

• The developed ANN model achieved a high correlation coefficient (R) of 0.956.
• The model demonstrated low relative error (RE) of 0.857 and root mean square error (RMSE) of 0.171.
• The prediction accuracy significantly surpassed previously reported results, indicating model robustness and feasibility.

Conclusions:

• ANN models incorporating the group contribution method provide satisfactory performance for predicting blood-brain barrier (BBB) penetration (logBB).
• The developed model offers a reliable and accurate method for predicting BBB permeability.
• This approach can aid in the early screening of drug candidates for CNS-targeted therapies.