Deep learning tools to accelerate antibiotic discovery.

Area of Science:

• Computational chemistry
• Machine learning in drug discovery
• Artificial intelligence for pharmaceuticals

Background:

• Machine learning (ML) and artificial intelligence (AI) are increasingly vital in drug discovery.
• Deep learning models efficiently explore high-dimensional data for designing compounds with specific properties, including antibacterial activity.

Purpose of the Study:

• This review examines deep learning frameworks for antibiotic discovery.
• It highlights physicochemical features and addresses dataset limitations in the field.

Main Methods:

• Covers discriminative models: convolutional neural networks, recurrent neural networks, graph neural networks.
• Discusses generative models: neural language models, variational autoencoders, generative adversarial networks, normalizing flow, and diffusion models.

Main Results:

• Deep learning offers efficient exploration of high-dimensional data for compound design.
• Identifies challenges in antimicrobial prediction: imbalanced data, limited datasets, validation, target strains, and structure.

Conclusions:

• Integrating deep generative models with bioinformatics, molecular dynamics, and data augmentation can overcome current challenges.
• These advancements hold the potential to enhance model performance and accelerate antimicrobial discovery.