DCPPS: Prediction of Kinase-Specific Phosphorylation Sites Using Dynamic Embedding and Cross-Representation...

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Kinase-specific phosphorylation is crucial for cellular signaling and regulation.
• Accurate prediction of phosphorylation sites aids protein function research, disease study, and drug development.
• Existing prediction methods often overlook critical positional and interaction features.

Purpose of the Study:

• To propose a novel computational model, DCPPS, for enhanced kinase-specific phosphorylation site prediction.
• To address limitations in feature representation by incorporating dynamic embeddings and cross-representation interactions.

Main Methods:

• Developed Dynamic Embedding Encoding (DEE) to capture amino acid semantics and positional information.
• Designed a Cross-Representation Interaction Unit (CRIU) to mine interactions between global and local sequence features.
• Integrated global protein information with DEE and CRIU for comprehensive feature extraction.

Main Results:

• DCPPS demonstrated superior predictive performance and scalability in kinase-specific phosphorylation prediction.
• Ablation studies confirmed the significant contribution of global protein information, DEE, and CRIU to prediction accuracy.
• The model effectively mitigated challenges posed by class imbalance in phosphorylation site prediction.

Conclusions:

• DCPPS offers a significant advancement in predicting kinase-specific phosphorylation sites.
• The integration of dynamic embeddings and cross-representation interactions enhances feature representation and predictive power.
• This model holds promise for accelerating research in signaling pathways, disease mechanisms, and therapeutic development.