Mitigating cold-start problems in drug-target affinity prediction with interaction knowledge transferring.

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning in drug discovery

Background:

• Drug-target interaction (DTI) prediction is vital for drug discovery and repurposing.
• Machine learning models for DTI often suffer from the cold-start problem with novel drugs or targets.
• Existing unsupervised learning methods for representation learning lack crucial interaction information.

Purpose of the Study:

• To address the cold-start problem in drug-target affinity (DTA) prediction.
• To incorporate essential interaction information into DTA models.
• To enhance the performance of machine learning models for DTI prediction.

Main Methods:

• Proposed a transfer learning approach.
• Utilized chemical-chemical interaction (CCI) and protein-protein interaction (PPI) datasets for pre-training.
• Transferred learned representations to the drug-target interaction (DTI) task.

Main Results:

• Demonstrated the effectiveness of transfer learning from CCI and PPI tasks.
• Showcased advantages over other pre-training methods in DTA.
• Improved prediction performance for novel drug-target pairs.

Conclusions:

• Transfer learning from related interaction tasks is a viable strategy to mitigate the cold-start problem in DTA.
• The proposed method offers a significant improvement for predicting drug-target interactions.
• This approach enhances the utility of machine learning in drug discovery and repurposing.