Analysis and prediction of human acetylation using a cascade classifier based on support vector machine.

Area of Science:

• Biochemistry
• Computational Biology
• Proteomics

Background:

• Lysine acetylation is a vital, reversible post-translational modification in biological processes.
• Accurate identification of acetylation sites is crucial for understanding protein function and mechanisms.
• Computational methods offer a faster and more convenient alternative to experimental techniques for site identification.

Purpose of the Study:

• To develop and validate a novel computational method for predicting human protein acetylation sites.
• To integrate diverse sequence and structural features for enhanced prediction accuracy.
• To address the challenge of imbalanced datasets in acetylation site prediction.

Main Methods:

• Combined sequence and structural features: physicochemical property (PCP), position-specific scoring matrix (PSSM), auto-covariance (AC), residue composition (RC), secondary structure (SS), and accessible surface area (ASA).
• Employed a two-step feature selection process combining minimum redundancy maximum relevance (mRMR) and iterative feature selection (IFS).
• Utilized a cascade classifier based on support vector machine (SVM) to handle sample imbalance and improve coverage.

Main Results:

• The proposed method achieved a specificity of 72.19% and a sensitivity of 76.71% on an independent dataset.
• Demonstrated that the cascade SVM classifier significantly outperforms a single SVM classifier.
• Successfully addressed the imbalance between positive and negative acetylation site samples.

Conclusions:

• The developed computational method provides an effective approach for predicting human acetylation sites.
• The cascade SVM classifier offers improved performance compared to traditional single SVM models.
• The study contributes a systematic analysis of acetylation data, enhancing our understanding of this modification.