Inter- and intra-chain disulfide bond prediction based on optimal feature selection

Area of Science:

• Computational Biology
• Protein Biochemistry
• Bioinformatics

Background:

• Protein disulfide bonds are vital post-translational modifications involved in physiological and pathological processes.
• Experimental determination of disulfide bonds is time-consuming and labor-intensive, especially for large datasets.
• Existing 3D structure-based prediction methods have limitations, necessitating sequence-based computational approaches.

Purpose of the Study:

• To develop a convenient and fast computational method for predicting both inter- and intra-chain protein disulfide bonds.
• To utilize sequence-based features for accurate disulfide bond prediction, overcoming limitations of experimental and 3D structure methods.

Main Methods:

• A computational method combining the maximum relevance and minimum redundancy (mRMR) technique with incremental feature selection (IFS).
• Nearest neighbor algorithm was employed as the prediction model.
• Features used include sequence conservation, residual disorder, amino acid factors, and sequential distance between cysteines.

Main Results:

• Achieved a prediction accuracy of 0.8702 for inter-chain disulfide bonds using 128 features.
• Achieved a prediction accuracy of 0.9219 for intra-chain disulfide bonds using 261 features.
• Identified key features and sites crucial for disulfide bond formation and revealed similarities/differences between inter- and intra-chain mechanisms.

Conclusions:

• The developed sequence-based computational method provides an efficient and accurate approach for predicting protein disulfide bonds.
• The findings offer insights into the mechanisms of disulfide bond formation, aiding further experimental studies.
• This method can accelerate research in areas involving protein structure, function, and disease.