Explainable machine learning predictions of dual-target compounds reveal characteristic structural features

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Multi-target compounds are crucial in modern drug discovery.
• Identifying structural features of these compounds remains a challenge.

Purpose of the Study:

• To leverage explainable machine learning (XML) to uncover structural motifs characteristic of dual-target compounds.
• To bridge the gap between predictive modeling and intuitive chemical understanding.

Main Methods:

• Development of accurate prediction models for dual-target activity using a specific test system.
• Quantification of molecular representation features to explain model predictions.
• Computational analysis of feature contributions to identify structural motifs.

Main Results:

• Identified specific molecular features distinguishing dual-target from single-target compounds.
• These features formed coherent substructures within dual-target molecules.
• Confirmed these substructures as signatures of specific dual-target activities.

Conclusions:

• Explainable machine learning effectively reveals characteristic substructures of dual-target compounds.
• This approach facilitates intuitive chemical analysis and aids in rational drug design.
• The identified motifs serve as valuable signatures for multi-target drug discovery.