CMV-GLA: contrastive multi-view graph layer attention for predicting phosphorylation site-disease associations

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Protein phosphorylation is a key posttranslational modification regulating cell signaling and metabolism.
• Dysregulation of phosphorylation is linked to various diseases, but phosphorylation site-disease associations are understudied.
• Existing prediction methods struggle to integrate diverse data, limiting performance.

Purpose of the Study:

• To develop a novel framework, CMV-GLA, for predicting phosphorylation site-disease associations.
• To leverage heterogeneous data sources including protein sequences, disease semantics, and known associations.
• To enhance predictive performance by integrating complementary information.

Main Methods:

• Developed Contrastive Multi-View Graph Layer Attention Network (CMV-GLA).
• Integrated protein sequences, disease semantics, and known associations into three graph views.
• Employed Graph Attention Network (GAT) backbone with layer attention and contrastive learning for robust feature representation.

Main Results:

• CMV-GLA significantly outperformed state-of-the-art methods in AUC and AUPRC on benchmark datasets.
• Ablation studies validated the importance of multi-view fusion and the contrastive learning module.
• Case studies provided high-confidence, literature-supported predictions.

Conclusions:

• CMV-GLA offers a powerful new approach for predicting phosphorylation site-disease associations.
• The framework enhances understanding of phosphorylation-mediated disease mechanisms.
• CMV-GLA can guide therapeutic discovery for phosphorylation-related diseases.