Accelerating De Novo Drug Design against Novel Proteins Using Deep Learning

Area of Science:

• Computational chemistry
• Drug discovery
• Artificial intelligence in medicine

Background:

• Emerging infectious diseases necessitate rapid drug design.
• Deep learning shows promise in drug design but struggles with novel targets due to data scarcity.
• Existing methods require substantial target-specific ligand data, hindering development for new diseases.

Purpose of the Study:

• To develop a deep learning and molecular modeling pipeline for accelerated drug design against novel targets with limited data.
• To create a robust drug design strategy applicable even without prior knowledge of target-specific ligands.
• To validate the pipeline's efficacy in designing novel, high-affinity inhibitors.

Main Methods:

• Screening homologous protein inhibitors against the target active site to generate an initial dataset.
• Employing transfer learning to extract features from the limited target-specific dataset.
• Utilizing a deep predictive model for docking score prediction and reinforcement learning to optimize molecular design.
• Validating the pipeline by designing inhibitors against human JAK2 without using existing inhibitors for training.

Main Results:

• The pipeline successfully generated an initial target-specific dataset using homologous inhibitors.
• Transfer learning effectively captured relevant molecular features from the limited dataset.
• The integrated deep learning model predicted docking scores and guided the design of new chemical entities.
• The approach reproduced known inhibitors and designed novel molecules with improved binding energy for human JAK2.

Conclusions:

• The developed drug design pipeline effectively addresses data scarcity challenges for novel targets.
• The combination of deep learning, molecular modeling, and reinforcement learning offers a powerful approach for accelerated drug discovery.
• The successful validation against human JAK2 demonstrates the method's potential for designing effective therapeutics against emerging diseases.