PTPPI: A Study on Protein Inhibitor Prediction Methods Using Multimodal Feature Fusion and Attention Mechanism.

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Protein-protein interactions (PPIs) are crucial for biological processes and disease pathogenesis.
• Developing small molecule inhibitors for PPIs is vital for treating diseases like cancer but is hampered by limited predictive data.
• Accurate prediction of novel PPI inhibitors requires advanced computational methods integrating multiple data types.

Purpose of the Study:

• To develop an efficient computational framework, PTPPI, for predicting protein-protein interaction inhibitors (PPIIs).
• To enhance the accuracy of inhibitor prediction by integrating diverse molecular features and employing advanced machine learning techniques.

Main Methods:

• PTPPI framework integrates extended connectivity fingerprints (ECFPs) and ChemBERTa-generated deep semantic embeddings of SMILES.
• Features are processed by independent encoders and fused using an interactive attention mechanism.
• A multi-task learning approach is used for feature reconstruction and inhibition score prediction.

Main Results:

• PTPPI demonstrated superior performance in inhibitor identification and potency prediction across eight PPI target families.
• The framework effectively integrates multiple molecular features, outperforming existing computational methods.
• Experimental validation confirmed PTPPI's reliability in predicting novel PPI inhibitors.

Conclusions:

• PTPPI offers a valuable and reliable tool for the discovery of novel protein-protein interaction inhibitors.
• The framework's ability to integrate diverse molecular data enhances prediction accuracy, advancing drug discovery efforts.
• PTPPI holds significant potential for developing new therapeutic strategies against diseases like cancer and viral infections.