iBhb-Lys: Identify lysine β-hydroxybutyrylation sites using autoencoder feature representation and fuzzy SVM...

Area of Science:

• Biochemistry
• Proteomics
• Computational Biology

Background:

• Lysine β-hydroxybutyrylation (Kbhb) is a novel histone modification linked to human disease pathogenesis.
• Accurate identification of Kbhb sites is crucial for understanding its biological significance and molecular mechanisms.

Purpose of the Study:

• To develop a novel computational model, iBhb-Lys, for accurate identification of Kbhb sites from protein sequences.
• To improve upon existing prediction methods for Kbhb site identification.

Main Methods:

• Combined four types of features to create a 3266-dimensional feature vector for each potential Kbhb site.
• Utilized an autoencoder network for dimensionality reduction of the high-dimensional feature space.
• Proposed a fuzzy support vector machine algorithm incorporating sample density to enhance classification robustness.

Main Results:

• The iBhb-Lys model achieved an AUC increase of 2.22% compared to the existing predictor KbhbXG on independent tests.
• Feature analysis identified the frequency of leucine and histidine residues near Kbhb sites as significant predictors.
• The developed model demonstrates superior performance in identifying Kbhb sites.

Conclusions:

• The iBhb-Lys model provides an effective computational tool for identifying lysine β-hydroxybutyrylation sites.
• Findings offer valuable insights into the molecular mechanisms underlying Kbhb and its role in diseases.
• This study contributes to the advancement of post-translational modification analysis in proteomics.