Prediction of cystine connectivity using SVM

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Disulphide bonds are crucial for protein structure and stability.
• Predicting disulphide bridges from amino acid sequences is challenging due to combinatorial complexity.

Purpose of the Study:

• To develop and evaluate a Support Vector Machine (SVM) model for predicting disulphide connectivity in proteins.
• To assess the model's performance with and without prior knowledge of cysteine bonding states.

Main Methods:

• A novel encoding scheme incorporating physico-chemical properties and statistical features was employed.
• Features included amino acid residue probabilities in secondary structures and PSI-blast profiles.
• The SVM model was trained and tested on the SPX dataset, derived from Swiss-Prot and PDB.

Main Results:

• The SVM model demonstrated robust performance in predicting cystine connectivity.
• Performance was evaluated against a recursive neural network model on the same dataset.
• The proposed encoding scheme effectively captures relevant sequence information for prediction.

Conclusions:

• The developed SVM model offers an effective approach for predicting disulphide bridges in protein sequences.
• The method provides a valuable tool for structural bioinformatics and protein engineering.
• Accurate prediction of disulphide bonds aids in understanding protein folding and function.