CNNArginineMe

CNNArginineMe predicts protein arginine methylation (PRme) sites using a Convolutional Neural Network (CNN) trained on One-Hot encoded peptide sequences to identify regulatory post-translational modification loci.

Key Features:

• Deep-Learning Architecture: Implements a Convolutional Neural Network (CNN) to detect sequence patterns associated with arginine methylation.
• One-Hot Encoding: Transforms peptide sequences into One-Hot encoded representations as input to the CNN.
• Performance Metrics: Evaluation uses Area Under the Curve (AUC) scores, with the CNN demonstrating superior AUC compared with existing PRme site predictors.

Scientific Applications:

• Predictive Modeling: Predicts arginine methylation sites across proteomes to map potential regulatory modification loci.
• Functional Analysis: Enables analysis of the arginine-methylated proteome and revealed significant enrichment in pathways associated with amyotrophic lateral sclerosis (ALS).

Methodology:

Multiple machine-learning models were constructed; the deep-learning model based on CNN architecture trained on One-Hot encoded peptide sequences achieved the best performance measured by AUC.