MANIFOLD: protein fold recognition based on secondary structure, sequence similarity and enzyme classification

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Science

Background:

• Accurate protein fold recognition is crucial for understanding protein function and biological processes.
• Existing sequence-based methods struggle with proteins exhibiting low sequence similarity to known structures.
• Integrating diverse structural and functional features can potentially improve fold recognition accuracy.

Purpose of the Study:

• To develop and evaluate a novel protein fold recognition method, MANIFOLD.
• To assess the contribution of predicted secondary structure, sequence, and enzyme code similarities in fold recognition.
• To investigate the impact of a non-linear ranking scheme for combining similarity measures.

Main Methods:

• MANIFOLD utilizes predicted secondary structure, sequence, and enzyme code similarities between target and template proteins.
• A non-linear ranking scheme was developed to optimally combine scores from these different similarity measures.
• Performance was evaluated on a challenging test set of proteins with limited sequence similarity.

Main Results:

• MANIFOLD achieved 34% accuracy in predicting the correct fold class for difficult test cases.
• This represents an over twofold increase in accuracy compared to sequence-based methods like PSI-BLAST and GenTHREADER (13-14% accuracy).
• Incorporating a functional similarity term further boosted prediction accuracy by up to 3%.

Conclusions:

• MANIFOLD demonstrates superior performance in protein fold recognition, particularly for proteins with low sequence similarity.
• Combining predicted secondary structure and functional information offers advantages beyond sequence similarity alone.
• The developed non-linear ranking scheme effectively integrates multiple similarity measures for improved prediction accuracy.