predML-Site: Predicting Multiple Lysine PTM Sites With Optimal Feature Representation and Data Imbalance Minimization

Area of Science:

• Computational biology
• Proteomics
• Biochemistry

Background:

• Post-translational modifications (PTM) significantly influence protein function and biological mechanisms.
• Accurate identification of multiple PTMs at lysine residues (K-PTM) is essential for advancing computational proteomics, cell biology, pathogenesis, and drug development.
• Existing computational methods for predicting K-PTM are still under development.

Purpose of the Study:

• To introduce predML-Site, a novel computational tool for predicting various K-PTMs, including acetylation, crotonylation, methylation, and succinylation.
• To develop a method capable of handling uncategorized peptide samples with single, multiple, or no modifications.
• To provide a user-friendly web-server for experimental scientists to facilitate K-PTM research.

Main Methods:

• Utilized multiple sequence encoding schemes (sequence-coupling, binary encoding, k-spaced amino acid pairs, amino acid factor) for informative feature representation.
• Employed ANOVA and incremental feature selection to identify optimal features.
• Adopted a cost-sensitive Support Vector Machine (SVM) classifier to address class-label imbalance.

Main Results:

• predML-Site achieved 84.18% accuracy in predicting multi-label PTM sites using the top 91 features.
• The tool demonstrated a high aiming rate of 85.34% and a coverage rate of 86.58%.
• Performance metrics significantly outperformed existing state-of-the-art predictors on the same validation dataset.

Conclusions:

• predML-Site is a highly accurate and effective computational tool for predicting multiple lysine post-translational modifications.
• The tool's superior performance and user-friendly web-server deployment offer valuable support for K-PTM studies.
• predML-Site has the potential to advance research in computational proteomics and related biological fields.