DeepSuccinylSite

DeepSuccinylSite predicts protein succinylation sites on lysine residues using deep learning applied to primary protein sequence to support study of this post-translational modification.

Key Features:

• Deep learning: Employs deep learning techniques to model complex sequence patterns associated with succinylation.
• Embedding strategies: Uses embedding strategies to represent primary protein sequence context for prediction.
• Sequence-based input: Predicts succinylation sites directly from primary protein sequence.
• Target residue and chemical effect: Focuses on succinylation of lysine residues and notes the succinyl group changes lysine net charge from +1 to −1 at physiological pH and can alter local protein structure.
• Validation and performance: Validated on an independent test set with sensitivity 79%, specificity 68.7%, Matthews correlation coefficient (MCC) 0.48, and area under the ROC curve 0.8.
• Compared performance: Demonstrates improved overall prediction accuracy relative to previously available succinylation predictors.

Scientific Applications:

• Site prioritization for experimental validation: Ranks candidate lysine succinylation sites to prioritize experimental follow-up.
• Functional analysis of PTMs: Supports investigation of succinylation effects on protein structure and function.
• Metabolism–PTM linkage studies: Enables studies linking succinate metabolism to succinylation and downstream cellular processes.
• Genomic regulation and disease research: Aids hypothesis generation regarding roles of succinylation in genomic regulation, DNA repair, and disease mechanisms.

Methodology:

Employs deep learning combined with embedding strategies on primary protein sequence and is validated on an independent test set (sensitivity 79%, specificity 68.7%, MCC 0.48, AUC 0.8).