Deciphering the functional landscape of phosphosites with deep neural network

Area of Science:

• Biochemistry
• Bioinformatics
• Systems Biology

Background:

• Current methods identify few kinases and phosphosites.
• Functional roles of most phosphosites remain unknown.

Purpose of the Study:

• To develop a novel deep neural network model, FuncPhos-SEQ, for functional assignment of human phosphosites.
• To analyze kinase substrate preferences and identify novel kinase-substrate interactions.

Main Methods:

• Integrated deep neural network (DNN) model incorporating convolutional neural network (CNN) channels for sequence motifs and DNN channels for protein-protein interaction (PPI) networks.
• Network embedding for PPI network features.
• In vitro and cellular biochemical assays.

Main Results:

• FuncPhos-SEQ successfully assigns functions to human phosphosites.
• NADK-S48/50 phosphorylation was confirmed to activate NADK enzymatic activity.
• ERK1/2 were identified as the primary kinases responsible for NADK-S48/50 phosphorylation.

Conclusions:

• FuncPhos-SEQ provides a powerful tool for phosphosite functional assignment.
• The study reveals a novel regulatory mechanism for NADK activity via ERK1/2 phosphorylation.
• FuncPhos-SEQ is available as an online server for broader research use.