Gaussian interaction profile kernels for predicting drug-target interaction.

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning

Background:

• Accurate prediction of drug-target interactions is crucial for drug discovery and target identification.
• Existing datasets contain limited experimentally validated drug-target pairs, necessitating advanced computational methods.
• Developing accurate computational tools is essential to bridge the gap between potential and validated drug-target interactions.

Purpose of the Study:

• To develop and evaluate a novel computational method for predicting drug-target interactions.
• To leverage network topology information for enhanced prediction accuracy.
• To assess the performance of the proposed method against existing state-of-the-art techniques.

Main Methods:

• Introduction of interaction profiles for drugs and targets as binary vectors representing network interactions.
• Definition of the Gaussian Interaction Profile (GIP) kernel based on these interaction profiles.
• Application of Regularized Least Squares (RLS) classifier with the GIP kernel for predicting drug-target interactions.
• Comparative analysis on established drug-target interaction networks.

Main Results:

• The proposed method, RLS with the GIP kernel, achieved a high Area Under the Precision-Recall Curve (AUPR) of up to 92.7.
• Significant improvement in prediction accuracy compared to current state-of-the-art methods was demonstrated.
• Incorporating chemical and genomic information alongside network topology further enhanced prediction accuracy, particularly for smaller datasets.

Conclusions:

• Network topology, represented by interaction profiles, is a highly relevant information source for predicting drug-target interactions.
• The developed computational method offers a powerful tool for identifying novel drug-target pairs.
• This approach holds significant promise for accelerating drug discovery and development pipelines.