M2PP: a novel computational model for predicting drug-targeted pathogenic proteins.

Area of Science:

• Biochemistry
• Computational Biology
• Drug Discovery

Background:

• Accurate detection of pathogenic proteins is crucial for understanding disease mechanisms and developing treatments.
• Genome-wide protein prediction is time-consuming and costly for targeted drug development.
• Predicting pathogenic proteins targeted by existing drugs can expedite disease treatment.

Purpose of the Study:

• To develop a computational model for predicting pathogenic proteins that are targeted by existing drugs.
• To facilitate faster and more cost-effective disease treatment strategies.

Main Methods:

• A novel computational model, M2PP, was developed.
• A heterogeneous network incorporating target proteins, diseases, and drugs was constructed.
• Features based on neighborhood similarity, drug inference, and path information were utilized.
• A random forest regression model was trained to score target-disease pairs.

Main Results:

• M2PP demonstrated superior prediction performance compared to state-of-the-art methods in five-fold cross-validation.
• The model accurately identified high-ranked pathogenic proteins for common diseases.
• Biomedical literature supported the model's predictions, highlighting its practical applicability.

Conclusions:

• M2PP is an effective and accurate tool for predicting drug-targeted pathogenic proteins.
• The model can significantly aid future biological research and drug development efforts.