Prediction of protein interactions with function in protein (de-)phosphorylation

Area of Science:

• Systems biology
• Network science
• Computational biology

Background:

• Protein-protein interactions (PPIs) form a complex cellular network known as the interactome.
• Evidence suggests hyperbolic geometry underlies complex network representations, including the human interactome (hPIN).
• Embedding the hPIN in hyperbolic space (H2) has previously captured biologically relevant information.

Purpose of the Study:

• To investigate if hyperbolic mapping of the hPIN can predict the function of PPIs, specifically those related to post-translational modifications (PTMs).
• To develop and evaluate a predictive model for PTM-related directed PPIs, focusing on protein phosphorylation and dephosphorylation.

Main Methods:

• Utilized a random forest algorithm to predict PTM-related directed PPIs.
• Employed hyperbolic properties and centrality measures derived from the hPIN mapped in H2 as features.
• Validated the algorithm by predicting PTM-related PPIs for ataxin-1, a protein implicated in Spinocerebellar Ataxia type 1 (SCA1).

Main Results:

• The algorithm successfully predicted PTM-related PPIs.
• Proteomics analysis in a cellular SCA1 model confirmed several predicted PTM-PPIs were dysregulated.
• Identified a compact cluster of ataxin-1, its dysregulated PTM-PPIs, and upstream regulators potentially critical for SCA1 pathology.

Conclusions:

• Hyperbolic network properties can be leveraged to predict PTM-related PPIs.
• The developed algorithm shows potential for inferring phosphorylation activity through directed PPIs.
• This approach may offer new insights into disease mechanisms involving PTMs and protein interactions.