Exploring Novel Fentanyl Analogues Using a Graph-Based Transformer Model.

Area of Science:

• Computational Chemistry
• Drug Discovery
• Artificial Intelligence

Background:

• Fentanyl and its analogues are easily modified, posing challenges for regulatory supervision due to limited database entries.
• Criminals exploit the structural variability of fentanyl analogues to evade detection.

Purpose of the Study:

• To develop a novel molecular generation model for creating new fentanyl analogues.
• To enhance the exploration of fentanyl molecule structures and understand their distribution patterns.

Main Methods:

• A molecular graph-based transformer model was employed.
• A data augmentation technique using substructure replacement was integrated.
• Generated molecules were screened, and their properties were calculated.

Main Results:

• 140,000 potential fentanyl analogues were generated, with 36,799 identified after screening.
• The model demonstrated an ability to learn properties of original fentanyl molecules.
• The proposed model generated more novel fentanyl analogues compared to other deep learning models.

Conclusions:

• The molecular graph-based transformer model effectively generates novel fentanyl analogues.
• This approach aids in exploring the structural landscape of potential fentanyl molecules.
• The findings contribute to understanding the distribution of fentanyl analogues.