Prediction of post-translational modification sites using multiple kernel support vector machine.

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Protein post-translational modifications (PTMs) are crucial for regulating protein function.
• Experimental identification of PTM sites is costly and labor-intensive.
• Computational methods using local sequence information and conserved motifs exist for PTM site prediction.

Purpose of the Study:

• To develop a novel computational method for predicting multiple PTM sites.
• To improve the accuracy and efficiency of PTM site prediction.
• To leverage local sequence information and site-modification relationships effectively.

Main Methods:

• Utilized a combination of multiple kernel support vector machines (SVM).
• Developed a local sequence kernel and a Gaussian interaction profile kernel.
• Integrated multiple kernels to train SVM for enhanced predictive performance.

Main Results:

• The proposed method demonstrated comparable or superior performance to existing PTM prediction methods.
• Experimental results validated the feasibility of the developed kernels.
• The method proved useful for predicting various PTM sites, including phosphorylation, O-linked glycosylation, acetylation, sulfation, and nitration.

Conclusions:

• The novel multiple kernel SVM approach is effective for PTM site prediction.
• The developed kernels efficiently utilize local sequence and interaction profile information.
• This method offers a valuable tool for computational PTM site identification.