DeepKme

DeepKme predicts lysine methylation sites in the human proteome using deep learning to identify post-translational modifications (PTMs) relevant to protein function and gene regulation.

Key Features:

• Deep Learning Architecture: Employs deep learning models to capture complex sequence patterns for accurate lysine methylation site prediction.
• Generalization Performance Assessment: Implements an experiment-split test that trains and tests on distinct experimental sources to assess model generalization beyond conventional cross-validation.
• Benchmarking Tool: Provides a practical performance measure for PTM prediction models that reflects generalization to unseen experimental data.

Scientific Applications:

• Proteomics: Identifies potential lysine methylation sites to inform analyses of protein function and protein–protein interactions.
• Epigenetics: Supports study of lysine methylation patterns that contribute to gene regulation mechanisms.
• Disease Research: Aids identification of aberrant methylation-related biomarkers in diseases including cancer.

Methodology:

Uses deep learning-based prediction and an experiment-split test that separates training and testing by experimental source to evaluate generalization versus conventional cross-validation.