A new representation for protein secondary structure prediction based on frequent patterns

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Bioinformatics

Background:

• Developing accurate protein secondary structure prediction methods is crucial for understanding protein function.
• Existing methods often rely on complex sequence alignments and evolutionary information.
• A new representation based on frequent amino acid patterns offers an alternative approach.

Purpose of the Study:

• To introduce and evaluate a novel representation for protein secondary structure prediction.
• To explore the use of frequent amino acid patterns and support vector machines (SVMs) for this task.
• To assess the performance of the new method against established prediction tools.

Main Methods:

• Identification of frequent amino acid patterns in protein sequences using a level-wise search.
• Definition of a feature set derived from these frequent patterns.
• Application of Support Vector Machines (SVMs) for secondary structure prediction using the defined features.

Main Results:

• The new representation, despite limited training data, achieved prediction accuracy comparable to leading methods like PSI-PRED and PROFsec.
• Evaluation was performed using 150 targets from the EVA contest in a blind testing setup.
• The proposed method demonstrated significant contribution to consensus predictions.

Conclusions:

• Frequent amino acid patterns provide a valuable feature set for protein secondary structure prediction.
• SVMs effectively utilize this new representation for accurate predictions.
• This method offers a promising alternative or complementary approach to existing secondary structure prediction techniques.