Probabilistic approach to predicting substrate specificity of methyltransferases

Area of Science:

• Biochemistry
• Computational Biology
• Enzymology

Background:

• Enzyme substrate specificity is crucial for biological function but challenging to predict.
• Existing methods often rely on species-specific data, limiting broad applicability.

Purpose of the Study:

• To develop a general, sequence-data-driven probabilistic framework for predicting enzyme substrate specificity.
• To apply this framework to yeast methyltransferases (MTases) and uncover novel substrate interactions.

Main Methods:

• Utilized a probabilistic model incorporating physico-chemical and biological properties (e.g., structural fold, isoelectric point, expression, localization).
• Employed Maximum Likelihood optimization and Akaike Information Criterion for model parameter tuning and variable selection.
• Focused on sequence-derived data for broad applicability across organisms and enzymes.

Main Results:

• The framework accurately predicted substrate specificity for yeast MTases, distinguishing between protein, RNA, and other substrates.
• Achieved 89% confirmation rate for experimental predictions.
• Identified YOR021C as the first known SPOUT-fold MTase with protein methylation activity, challenging existing classifications.

Conclusions:

• The developed framework offers a powerful and generalizable tool for predicting enzyme substrate specificity.
• Provides insights into the general principles governing methyltransferase substrate recognition.
• Advances our understanding of enzyme function and evolution through accurate functional prediction.