Addressing the Problem of Lysine Glycation Prediction in Proteins via Recurrent Neural Networks

Area of Science:

• Biochemistry
• Molecular Biology
• Computational Biology

Background:

• Diabetes mellitus is a metabolic disorder characterized by cellular damage.
• Glycation, a nonenzymatic reaction between sugars and biomolecules, forms detrimental advanced glycation end-products (AGEs).
• Identifying glycation sites on proteins, particularly lysine residues, is challenging due to the lack of clear sequence motifs.

Purpose of the Study:

• To investigate the influence of physicochemical properties on lysine glycation.
• To develop a predictive model for identifying potential lysine glycation sites in proteins.
• To enhance understanding of glycation mechanisms.

Main Methods:

• Utilized a curated CPLM database.
• Implemented a recurrent neural network strategy for classification.
• Evaluated the predictive power of amino acid physicochemical properties, including isoelectric point, mass, and torsion angle.

Main Results:

• Achieved 59.6% accuracy in predicting lysine glycation using the isoelectric point of flanking amino acids.
• Increased prediction accuracy to approximately 60% by combining mass and torsion angle properties.
• Demonstrated that physicochemical properties of neighboring amino acids influence glycation site selection.

Conclusions:

• Physicochemical properties, particularly isoelectric point, mass, and torsion angle, play a significant role in determining lysine glycation.
• The developed recurrent neural network approach aids in predicting and understanding glycation principles.
• This work provides a valuable tool for identifying potential lysine glycation sites for further experimental investigation.