HybridSucc

HybridSucc predicts lysine succinylation (Ksucc) sites across multiple species with emphasis on human-specific predictions to identify candidate modification sites for biological and disease-related studies.

Key Features:

• Hybrid-Learning Architecture: Integrates deep learning with conventional machine learning algorithms within a unified framework to improve prediction accuracy.
• Comprehensive Dataset: Built on 26,243 non-redundant known Ksucc sites from 13 species used for training and validation.
• Informative Feature Integration: Employs 10 types of informative features to capture patterns that distinguish succinylated from non-succinylated lysine residues.
• High Predictive Accuracy: Achieves area under curve (AUC) values of 0.885 for general prediction and 0.952 for human-specific prediction, representing 17.84% to 50.62% improvements over existing tools.
• Proteome-Wide Prediction Capability: Enables proteome-wide identification of potential Ksucc sites and prioritization of mutations that alter succinylation states in proteins such as PKM2, SHMT2, and IDH2.
• Research-Grade Candidate Generation: Provides candidate Ksucc sites for experimental validation to support studies of protein acylation in biological processes and disease mechanisms.

Scientific Applications:

• Proteome-wide Ksucc Prediction: Enables identification of potential succinylation sites across entire proteomes.
• Prioritization of Cancer-Related Mutations: Facilitates ranking of mutations that may alter Ksucc states in proteins including PKM2, SHMT2, and IDH2.
• Experimental Candidate Generation: Produces candidate sites for experimental validation of lysine succinylation.
• Investigation of Disease Mechanisms: Supports studies on the role of lysine succinylation in biological processes and the pathogenesis of human cancers.

Methodology:

HybridSucc applies a hybrid-learning architecture combining deep learning and conventional machine learning, trained and validated on a dataset of 26,243 non-redundant Ksucc sites from 13 species using 10 types of informative features.