Development of a 2D-QSAR Model for Tissue-to-Plasma Partition Coefficient Value with High Accuracy Using Machine...

Area of Science:

• Pharmacokinetics and Drug Development
• Computational Chemistry
• Quantitative Structure-Activity Relationship (QSAR) studies

Background:

• Physiologically based pharmacokinetic (PBPK) models are increasingly used in new drug applications.
• Tissue-to-plasma partition coefficient (Kp) is a critical PBPK parameter, but experimental determination is challenging.
• In silico methods are sought to overcome experimental limitations in Kp determination.

Purpose of the Study:

• To develop a 2D-QSAR model for predicting Kp values.
• To utilize physicochemical descriptors as explanatory variables, moving beyond in vitro and in vivo parameters.
• To enhance the predictability of Kp values through an improved modeling approach.

Main Methods:

• A 2D-QSAR model was developed using the random forest algorithm.
• Physicochemical descriptors were employed as explanatory variables.
• A multimodal approach combined minimal in vitro experimental data with physicochemical descriptors.

Main Results:

• The developed multimodal 2D-QSAR model demonstrated superior accuracy compared to conventional models.
• The model achieved the highest accuracy, indicated by a low RMSE and high percentage of two-fold error (0.39 and 64.5%).
• Multimodality was confirmed as a beneficial strategy for 2D-QSAR model development.

Conclusions:

• A highly accurate 2D-QSAR model for Kp prediction was successfully developed.
• The model effectively integrates limited in vitro data with physicochemical descriptors.
• This approach offers a promising alternative for estimating Kp values in drug development.