Machine learning-driven prediction of substrates for enzymes introducing or removing protein post-translational...

Area of Science:

• Biochemistry
• Proteomics
• Computational Biology

Background:

• Post-translational modifications (PTMs) are crucial for understanding cellular processes, disease mechanisms, and cancer therapeutics.
• Identifying specific enzyme-mediated PTM sites is challenging due to the complexity of the proteome.

Purpose of the Study:

• To develop and validate a machine learning (ML)-driven method for predicting novel enzyme-specific PTM sites.
• To enhance the discovery of enzyme-substrate networks involved in PTM pathways.

Main Methods:

• Integration of machine learning with enzyme-mediated modification of complex peptide arrays.
• Experimental validation using mass spectrometry to confirm predicted PTM sites and substrate dynamics.
• Analysis of enzyme function in disease contexts, such as breast cancer mutations.

Main Results:

• The ML approach successfully predicted 37-43% of proposed PTM sites, identifying novel substrates for SET8 and SIRT1-7.
• Demonstrated significant performance improvement over traditional in vitro methods.
• Confirmed dynamic methylation for SET8 substrates and deacetylation of 64 unique sites for SIRT2.
• Revealed altered SET8 substrate networks in breast cancer missense mutations, highlighting differential enzyme function in disease.

Conclusions:

• The developed ML method offers a powerful tool for uncovering unexplored PTM sites and enzyme-substrate interactions.
• This approach provides valuable insights into enzyme specificity and differential enzyme function in disease.
• Potential for broad application in dissecting complex PTM pathways and identifying therapeutic targets.