OLB-AC: toward optimizing ligand bioactivities through deep graph learning and activity cliffs.

Area of Science:

• Computational chemistry and cheminformatics
• Drug discovery and development
• Machine learning and artificial intelligence in chemistry

Background:

• Deep graph learning (DGL) is crucial for ligand-based virtual screening.
• Activity cliffs (ACs) pose a challenge, as small molecular changes drastically alter bioactivity.
• Existing DGL models improve prediction near ACs, but optimization opportunities remain underexplored.

Purpose of the Study:

• To develop a novel deep graph learning approach for simultaneous prediction and optimization of ligand bioactivities near activity cliffs.
• To introduce a method that directly optimizes ligand molecules, providing a reference for enhancing bioactivity.
• To explore the potential of activity cliffs for optimizing ligand bioactivity in drug discovery.

Main Methods:

• Proposed a novel approach named OLB-AC (Optimizing Ligand Bioactivities near Activity Cliffs) utilizing deep graph learning.
• Developed an attentive graph reconstruction neural network to reconstruct and optimize ligands.
• Employed adversarial representations derived from bioactivity prediction gradients for ligand optimization.

Main Results:

• OLB-AC successfully optimized 667 molecules, identifying 49 known highly active/inhibitor/non-toxic ligands beyond the training datasets.
• Generated 27 novel molecular pairs with transformations not present in the training sets.
• Achieved state-of-the-art performance in bioactivity prediction, showing the best Pearson correlation coefficient (r2) on 27/33 datasets, with 7.2%-22.9% improvement.

Conclusions:

• The OLB-AC method effectively optimizes ligand bioactivities near activity cliffs, demonstrating significant potential in drug discovery.
• The approach generates novel molecular transformations and improves bioactivity prediction accuracy.
• Code and datasets are publicly available, facilitating further research in DGL for drug optimization.