Accurate prediction of kinase-substrate networks using knowledge graphs

Area of Science:

• Molecular Biology
• Computational Biology
• Bioinformatics

Background:

• Protein phosphorylation is crucial for cell signaling and fate.
• Identifying kinase-substrate interactions is vital but challenging.
• Existing computational methods have limited kinome coverage and accuracy.

Purpose of the Study:

• To develop an improved computational model for predicting kinase-substrate relationships.
• To enhance the accuracy and coverage of kinase-substrate predictions.
• To facilitate the discovery of novel phosphorylation events.

Main Methods:

• Utilized statistical relational learning on phosphorylation networks.
• Interpreted phosphorylation networks as knowledge graphs.
• Compared the novel model against six existing prediction systems.

Main Results:

• The developed model demonstrated superior kinome coverage compared to existing tools.
• Generated high-confidence, biologically valid predictions.
• Experimentally validated novel phosphorylations by LATS1, AKT1, PKA, and MST2 kinases.

Conclusions:

• The novel predictive model significantly advances kinase-substrate relationship discovery.
• The tool effectively focuses phosphoproteomic experiments.
• Facilitates the identification of previously unknown phosphorylation reactions and kinases.