What Works, What Doesn't, and Why? An Industrial Perspective on Absorption Modeling

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Pharmacokinetics

Background:

• Lead optimization failures in drug discovery are frequently attributed to poor absorption, efflux transport, and low recovery rates.
• Existing models often lack comprehensive analysis of industrial chemical spaces and transport route characterization.

Purpose of the Study:

• To develop a comprehensive model for predicting organic molecule absorption using public and industrial data.
• To identify key parameters influencing multiparameter optimization (MPO) and improve drug discovery efficiency.

Main Methods:

• Comprehensive modeling of public and industrial data on organic molecule absorption.
• Comparative analysis of industrial chemical space to examine permeability parameters.
• Multitask learning approach for model development and validation.

Main Results:

• Demonstrated the critical importance of recovery, distribution coefficient, and topological polar surface area in MPO.
• Revealed discrepancies between public and industrial data due to protocol variations, highlighting the need for standardized measurements.
• Identified misconceptions in transport route characterization within industrial chemical spaces.

Conclusions:

• Industrial data is crucial for avoiding applicability domain issues and ensuring measurement standardization in predictive modeling.
• Coupling predictive models with generative topographic mapping provides a visual strategy for chemical space exploration and optimization.
• The proposed approach supports MPO and aims to enhance overall drug discovery efficiency.