Nphos: Database and Predictor of Protein N-phosphorylation.

Area of Science:

• Biochemistry
• Proteomics
• Bioinformatics

Background:

• Protein N-phosphorylation is a crucial post-translational modification with limited research compared to O-phosphorylation.
• Understanding N-phosphorylation is vital for elucidating diverse biological processes.

Purpose of the Study:

• To establish a comprehensive database (Nphos) of experimentally verified protein N-phosphorylation sites.
• To characterize the features of N-phosphorylation and develop predictive models.
• To identify novel N-phosphosites in the human proteome.

Main Methods:

• Collected and curated 11,710 N-phosphosites from 7,344 proteins across 39 species.
• Analyzed sequential and structural features of N-phosphosites.
• Developed and optimized Gradient Boosting Decision Tree (GBDT) models for predicting pHis, pLys, and pArg.

Main Results:

• Constructed the Nphos database, providing a valuable resource for N-phosphorylation research.
• Achieved high prediction accuracy for human N-phosphorylation: AUCs of 90.56% (pHis), 91.24% (pLys), and 92.01% (pArg).
• Discovered 488,825 novel N-phosphosites in the human proteome.

Conclusions:

• The Nphos database and predictive models offer a robust foundation for N-phosphorylation research.
• This work significantly deepens the systematic understanding of protein N-phosphorylation.
• Facilitates future investigations into the roles of N-phosphorylation in biological systems.