Quantitative structure-pharmacokinetic relationship modelling: apparent volume of distribution

Area of Science:

• Pharmacokinetics
• Medicinal Chemistry
• Computational Chemistry

Background:

• The apparent volume of distribution (Vd) is a key pharmacokinetic parameter influencing drug dosing and efficacy.
• Predicting Vd accurately is crucial for effective drug development and therapeutic drug monitoring.
• Existing methods for Vd prediction often lack accuracy for heterogeneous drug sets.

Purpose of the Study:

• To develop a quantitative structure-activity relationship (QSAR) model for predicting the apparent volume of distribution (Vd) in humans.
• To investigate the relationship between structural descriptors and Vd, including protein-binding corrected Vd (unbound Vd).
• To assess the predictive performance of QSAR models for Vd and unbound Vd across a diverse range of drugs.

Main Methods:

• Utilized stepwise regression analysis to build QSAR models.
• Included 70 drugs in the dataset, encompassing both acidic and basic compounds.
• Employed a test set comprising half the chemicals to evaluate model predictive power.

Main Results:

• QSAR models for Vd demonstrated lower prediction errors compared to models for unbound Vd (mean fold error of 2.01 vs. 2.28).
• Separating drugs into acidic and basic categories did not significantly improve prediction accuracy.
• The developed QSAR models showed robust predictive capabilities for the apparent volume of distribution.

Conclusions:

• A reliable QSAR model for predicting human drug Vd was successfully developed.
• The model's predictive performance is superior for Vd compared to unbound Vd.
• The findings suggest that a single QSAR model can effectively predict Vd for a heterogeneous collection of drugs.