Contribution assessment of multiparameter optimization descriptors in CNS penetration

Area of Science:

• Medicinal Chemistry
• Pharmacokinetics
• Computational Chemistry

Background:

• Predicting blood-brain barrier (BBB) penetration is crucial for developing central nervous system (CNS) drugs.
• The multiparameter optimization (MPO) approach is a common strategy for assessing drug properties.
• However, the MPO approach's effectiveness for CNS/non-CNS classification requires further investigation.

Purpose of the Study:

• To evaluate the influence of six physicochemical properties on BBB penetration using logistic and multiple linear regression.
• To explore alternative quantitative structure-activity relationship (QSAR) methods for CNS penetration prediction.
• To develop a reliable method for assessing the pharmacokinetic parameter log C_u,b.

Main Methods:

• Logistic regression and multiple linear regression were applied to a dataset of 578 chemicals.
• Quantitative structure-activity relationship (QSAR) analysis was performed using linear discriminant analysis, random forest, support vector machine, and Gaussian process.
• Subset clustering and graphic comparison of experimental and calculated values were used to analyze different penetration mechanisms.

Main Results:

• The aggregated MPO-score descriptor provided unsatisfactory results for CNS/non-CNS classification.
• Three properties—molecular weight, number of H-bond donors, and octanol-water partition coefficient—yielded optimal categorical models with modest accuracy (0.730-0.765).
• Regression models for specific subsets showed satisfactory statistics, enabling the estimation of descriptor contributions to log C_u,b.

Conclusions:

• The MPO approach for CNS penetration is ambiguous and requires refinement.
• QSAR analysis offers a more robust alternative for predicting BBB penetration.
• The developed additive scheme allows for preliminary quantitative assessment of log C_u,b by medicinal chemists.