Drug-Target Interaction Prediction Based on Multisource Information Weighted Fusion

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Existing drug-target interaction studies often assume no interaction for unknown relationships.
• This assumption overlooks potential interactions that are merely unconfirmed.

Purpose of the Study:

• To propose a weighted fusion method for screening drug-target interactions.
• To improve the accuracy and performance of drug-target interaction prediction models.

Main Methods:

• Treating unknown drug-target relationships as unlabeled samples.
• Selecting potential interacting drug-target pairs and adding them to positive samples.
• Revising the interaction matrix with new positive samples.
• Utilizing a bipartite local model with neighbor-based interaction profile inferring (BLM-NII) on revised datasets.

Main Results:

• The proposed method significantly improves specificity, sensitivity, precision, and accuracy compared to the standard BLM-NII.
• The method achieves excellent performance in Area Under the ROC Curve (AUC) and Area Under the Precision-Recall Curve (AUPR) compared to state-of-the-art methods.

Conclusions:

• The weighted fusion method effectively handles unlabeled data in drug-target interaction prediction.
• This approach offers a significant advancement in predicting drug-target relationships, crucial for drug discovery.