Related Experiment Video
Updated: Aug 10, 2026
09:40
A Mass Spectrometry-Based Proteomics Approach for Global and High-Confidence Protein R-Methylation Analysis
Published on: April 28, 2022
PLM-ArgMe: Protein language model for arginine methylation prediction for different species
Nitika Bhatt1, Kartik Joshi1, Ranjeet Kumar Rout2
1Center of Artificial Intelligence, Dr. B R Ambedkar National Institute of Technology Jalandhar, Jalandhar, Punjab, 144008, India.
Biochemical and Biophysical Research Communications
|August 8, 2026
Summary
Accurate prediction of protein methylation sites is vital for understanding disease mechanisms. The novel PLM-ArgMe model enhances prediction accuracy by integrating evolutionary, structural, and biochemical features using a symmetry-sensitive Transformer framework.
Area of Science:
- Biochemistry and Molecular Biology
- Computational Biology and Bioinformatics
- Genomics and Proteomics
Background:
- Protein methylation is a key post-translational modification (PTM) implicated in various diseases.
- Accurate prediction of methylation sites is crucial for elucidating disease molecular mechanisms.
- Existing deep learning models often rely solely on sequence data or handcrafted features, limiting their ability to capture complex biological properties and cross-species generalization.
Purpose of the Study:
- To develop an advanced computational framework for accurate prediction of arginine methylation sites.
- To address the limitations of existing methods by incorporating biochemical properties and positional patterns.
- To improve cross-species generalization and prediction accuracy in methylation site identification.
Main Methods:
- Development of PLM-ArgMe, a symmetry-sensitive Transformer framework.
- Utilizing context-aware ESM-2 residue embeddings for evolutionary and structural context.
- Implementation of a novel Bio-Symmetric Mirrored Sinusoidal Encoding (BSMSE) strategy to capture biological symmetry.
- Integration of physicochemical features to enhance biochemical representations.
- Employing symmetry-aware positional encoding and bidirectional multi-head self-attention to model dependencies.
Main Results:
- PLM-ArgMe achieved high prediction accuracies: 90.91% (Chimpanzee), 93% (Rat), 87.44% (Human), and 87.22% (Mouse).
- The model attained an overall accuracy of 88.41% when trained and evaluated on a combined multi-species dataset.
- Demonstrated robust generalization capabilities across diverse datasets.
Conclusions:
- PLM-ArgMe represents a significant advancement in arginine methylation site prediction.
- The framework's ability to integrate diverse biological features and address symmetry enhances prediction accuracy and generalization.
- The proposed method offers a robust tool for advancing research in PTMs and related diseases.
