pSuc-FFSEA: Predicting Lysine Succinylation Sites in Proteins Based on Feature Fusion and Stacking Ensemble Algorithm

Area of Science:

• Biochemistry
• Proteomics
• Bioinformatics

Background:

• Protein succinylation is a critical post-translational modification influencing protein structure and cellular functions.
• Dysregulation of succinylation is linked to various human diseases, necessitating efficient identification of succinylation sites.
• Experimental methods for identifying succinylation sites are laborious and cannot keep pace with growing biological datasets.

Purpose of the Study:

• To develop an accurate and efficient computational tool for predicting lysine succinylation sites in protein sequences.
• To overcome the limitations of experimental methods and existing prediction tools for succinylation site identification.

Main Methods:

• Feature extraction using EBGW, One-Hot, continuous bag-of-words, chaos game representation, and AAF_DWT.
• Feature selection via LASSO and construction of a two-layer stacking ensemble classifier (SVM, Broad Learning System, LightGBM, Logistic Regression).
• Hyperparameter optimization using Bayesian and grid search algorithms.

Main Results:

• The pSuc-FFSEA predictor achieved an average accuracy of 0.7773 ± 0.0120.
• Demonstrated excellent robustness and superior performance compared to existing prediction tools.
• A user-friendly web server (https://bio.cangmang.xyz/pSuc-FFSEA) was developed for easy access by researchers.

Conclusions:

• pSuc-FFSEA provides a robust and accurate method for predicting lysine succinylation sites.
• The developed tool facilitates research into the mechanisms of succinylation and its role in diseases.
• The accessible web server lowers the barrier for experimental scientists to utilize computational predictions.