A Machine Learning-Based Biological Drug-Target Interaction Prediction Method for a Tripartite Heterogeneous Network.

Area of Science:

• Pharmaceutical Sciences
• Computational Biology
• Machine Learning

Background:

• Drug repositioning accelerates pharmaceutical development by identifying new uses for existing drugs.
• Traditional experimental validation is costly and time-consuming.
• Computational methods, particularly kernel methods, are increasingly used for drug-target interaction prediction.

Purpose of the Study:

• To develop a novel machine learning prediction method for drug-target interactions.
• To integrate multiple biological data sources simultaneously using a tripartite heterogeneous network.
• To improve the accuracy and efficiency of computational drug repositioning.

Main Methods:

• A new machine learning algorithm combining multiple kernels.
• Extension of bipartite drug-target graphs to tripartite heterogeneous drug-target-disease interaction spaces.
• Utilizing Gaussian kernel functions and regularized least squares with Kronecker products for prediction.

Main Results:

• The proposed algorithm significantly improved prediction performance.
• Achieved high Area Under the Precision-Recall Curve (AUPR) and Area Under the Receiver Operating Characteristic Curve (AUC) values.
• AUC values reached 0.99, 0.99, 0.97, and 0.96 on four benchmark datasets, demonstrating superior accuracy.

Conclusions:

• The novel network algorithm effectively predicts drug-target interactions by integrating diverse biological information.
• Network topology is a valuable feature for predicting drug-target interactions.
• This approach offers a more efficient and cost-effective alternative to traditional drug development.