Prediction of protein secondary structure using probability based features and a hybrid system

Area of Science:

• Biochemistry
• Computational Biology
• Structural Biology

Background:

• Protein secondary structure prediction is crucial for understanding protein function.
• Existing methods often rely on complex sequence profiles.
• Developing accurate single-sequence prediction methods remains a challenge.

Purpose of the Study:

• To propose novel co-occurrence probability-based features for protein secondary structure prediction.
• To evaluate the performance of these features using a hybrid NEUROSVM classifier.
• To compare the proposed single-sequence methods (NSVMps and NSVM) against existing approaches.

Main Methods:

• Extraction of position-specific and position-independent co-occurrence probability features from protein sequences.
• Utilizing a hybrid NEUROSVM classifier combining neural networks and support vector machines.
• Testing on benchmark datasets (RS126, CB513) and CASP9 targets.

Main Results:

• The proposed features and NEUROSVM classifier outperform many existing single-sequence prediction methods.
• NSVMps, using lower-dimensional features, shows comparable results to other established methods.
• Performance was validated on CASP9 targets and compared with recent single-sequence predictors.

Conclusions:

• Co-occurrence probability-based features are effective for single-sequence protein secondary structure prediction.
• The NSVMps method offers a computationally efficient yet accurate alternative.
• Further investigation into amino acid residue contributions to secondary structure formation is warranted.