Deep_KsuccSite

Deep_KsuccSite predicts lysine succinylation (Ksucc) sites within protein sequences to identify post-translational modification locations relevant to protein function and regulation.

Key Features:

• Deep Learning Framework: Employs deep learning algorithms to model sequence patterns associated with Ksucc sites.
• Feature Encoding Techniques: Encodes peptide sequences using Composition, Transition, and Distribution (CTD) Composition (CTDC), Enhanced Grouped Amino Acid Composition (EGAAC), and Amphiphilic Pseudo-Amino Acid Composition (APAAC).
• Model Architecture: Constructs three base classifiers that include one-dimensional convolutional neural networks (1D-CNN) and two-dimensional convolutional neural networks (2D-CNN).
• Voting Ensemble: Integrates outputs from multiple classifiers using a voting method to produce final predictions.
• Validation and Benchmarking: Evaluates model performance using K-fold cross-validation, independent testing, and ablation experiments to assess feature combinations and architecture choices.
• Demonstrated Performance: Ablation experiments and benchmarking indicate improved prediction accuracy through the combined use of diverse feature encodings and ensemble voting.

Scientific Applications:

• Proteome-wide Ksucc annotation: Enables large-scale prediction of lysine succinylation sites across protein sequences.
• Functional regulation studies: Supports investigation of how succinylation affects protein function and regulation.
• Metabolic and disease research: Facilitates exploration of metabolic pathways and disease mechanisms associated with aberrant succinylation patterns.

Methodology:

Peptide sequences are encoded with CTDC, EGAAC, and APAAC; deep learning models (1D-CNN and 2D-CNN) are trained to form three base classifiers whose outputs are combined by a voting method; performance is assessed via K-fold cross-validation, independent testing, and ablation experiments.