A Quantitative Structure-Activity Relationship for Human Plasma Protein Binding: Prediction, Validation and...

Area of Science:

• Computational chemistry
• Pharmacology
• Drug discovery

Background:

• Plasma protein binding is a critical factor influencing drug efficacy and pharmacokinetics.
• Accurate prediction of plasma protein binding is essential for effective drug development.
• Existing methods for assessing plasma protein binding can be time-consuming and resource-intensive.

Purpose of the Study:

• To develop a robust and externally predictive in silico Quantitative Structure-Activity Relationship (QSAR)-neural network model.
• To predict the plasma protein binding of drugs.
• To enhance drug discovery by minimizing chemical synthesis and laboratory testing.

Main Methods:

• A dataset of 277 drugs was utilized for model development.
• A Filter method was employed to select 55 relevant molecular descriptors.
• The model's external predictive accuracy was evaluated using Q2 and RMSE metrics.

Main Results:

• The developed QSAR-neural network model demonstrated robustness and a good applicability domain.
• High external accuracy was achieved on the validation set, with Q2 = 0.966 and RMSE = 0.063.
• The model outperformed previously published models in predicting plasma protein binding.

Conclusions:

• An advanced QSAR-neural network model for predicting human plasma protein binding was successfully developed.
• The model accurately predicts plasma protein binding for a diverse set of 277 drugs.
• This tool can significantly aid drug discovery by reducing the reliance on extensive experimental validation.