Leveraging Network Target Theory for Efficient Prediction of Drug-Disease Interactions: A Transfer Learning Approach

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug discovery relies on efficient methods to identify potential therapeutics.
• Predicting drug-disease interactions is crucial for developing innovative treatments.
• Existing models face challenges in handling large datasets and balancing sample types.

Purpose of the Study:

• To develop a novel transfer learning model for predicting drug-disease interactions.
• To leverage network target theory and deep learning for enhanced drug feature extraction.
• To identify novel synergistic drug combinations for specific diseases, including cancer.

Main Methods:

• Utilized a transfer learning model integrating deep learning with biological molecular networks.
• Employed network techniques to extract precise drug features from existing knowledge.
• Addressed the challenge of balancing large-scale positive and negative samples in the dataset.

Main Results:

• Identified 88,161 drug-disease interactions involving 7,940 drugs and 2,986 diseases.
• Achieved an Area Under Curve (AUC) of 0.9298 and an F1 score of 0.6316 for interaction prediction.
• Successfully predicted drug combinations with an F1 score of 0.7746 and identified two novel synergistic combinations for cancer.

Conclusions:

• The novel model significantly improves the prediction of drug-disease interactions and drug combinations.
• The identified synergistic drug combinations show potential for effective cancer treatment regimens.
• This approach offers a powerful tool for accelerating drug development and personalized medicine.