Estimation of ADME properties with substructure pattern recognition.

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• Drug discovery relies heavily on evaluating absorption, distribution, metabolism, and excretion (ADME) properties.
• In vivo and in vitro ADME assessments are resource-intensive, driving the need for efficient in silico methods.
• Traditional in silico approaches often depend on molecular descriptors, raising concerns about their accuracy and efficiency.

Purpose of the Study:

• To develop a novel classification method for predicting ADME properties using substructure pattern recognition.
• To establish a direct link between molecular substructures and key ADME properties.
• To identify significant substructure patterns for medicinal chemistry interpretation.

Main Methods:

• Molecules represented as substructure pattern fingerprints based on a predefined dictionary.
• Support Vector Machine (SVM) algorithm employed for building predictive models.
• Information gain analysis utilized to identify important substructure patterns.

Main Results:

• High predictive accuracies achieved: 98.5% for human intestinal absorption (HIA) and 98.8% for blood-brain barrier (BBB) penetration on training sets.
• Excellent performance on test sets with 98.8% for HIA and 98.4% for BBB penetration.
• Identification of key substructure patterns significantly correlated with HIA and BBB penetration properties.

Conclusions:

• The proposed substructure pattern recognition method offers a reliable and accurate approach for in silico ADME property prediction.
• This method provides valuable insights into structure-property relationships, aiding medicinal chemistry efforts.
• The findings support the utility of this novel approach in accelerating drug discovery and development.