KenDTI: An Ensemble Model for Predicting Drug-Target Interaction by Integrating Multi-Source Information.

Area of Science:

• Computational Biology
• Bioinformatics
• Drug Discovery

Background:

• Drug-target interaction (DTI) identification is crucial for drug discovery but experimental validation is costly and inefficient.
• Computational models are vital for predicting potential DTIs, with performance depending on feature extraction from drugs and proteins.

Purpose of the Study:

• To develop an advanced computational model for accurate DTI prediction by combining diverse data sources and methodologies.
• To improve DTI prediction performance by leveraging both network-based and classification-based approaches.

Main Methods:

• Exploited biochemical drug characteristics using network integration.
• Utilized molecular sequences via word embeddings for feature representation.
• Developed an ensemble model, KenDTI, integrating network-based and classification-based methods.

Main Results:

• KenDTI demonstrated superior performance compared to state-of-the-art DTI predictors on large-scale datasets.
• The model showed robustness against missing network data and limited prior knowledge.
• Successfully predicted DTIs for drug candidates with scarce information.

Conclusions:

• Ensemble modeling integrating network and sequence data significantly enhances DTI prediction accuracy.
• KenDTI offers a robust and versatile tool for drug discovery, particularly for compounds with limited data.
• The developed approach advances computational drug discovery by improving the efficiency and reliability of identifying potential drug-target interactions.