An ensemble approach to protein fold classification by integration of template-based assignment and support vector...

Area of Science:

• Computational Biology
• Bioinformatics
• Structural Bioinformatics

Background:

• Protein fold classification is essential for protein structure prediction.
• Existing methods include template-based fold assignment and ab-initio prediction.
• Combining these approaches for enhanced accuracy was previously unexplored.

Purpose of the Study:

• To develop and evaluate a novel ensemble method for protein fold classification.
• To combine template-based and ab-initio algorithms to improve prediction accuracy.
• To assess the performance of the ensemble method against existing approaches.

Main Methods:

• Developed HH-fold, a template-based fold assignment algorithm using HHsearch.
• Developed SVM-fold, an ab-initio classification algorithm using support vector machines and sequence profiles.
• Combined HH-fold and SVM-fold into an ensemble method named TA-fold.

Main Results:

• TA-fold achieved 0.799 accuracy on the DD dataset (27 folds), improving over ab-initio methods by 5.4-11.7%.
• Accuracy increased to 0.971 on an updated DD dataset with more proteins.
• TA-fold demonstrated >0.9 accuracy on a large dataset (6451 proteins, 184 folds) and outperformed other threading methods on the LE dataset.

Conclusions:

• The ensemble approach TA-fold effectively combines template-based and ab-initio methods for superior protein fold classification.
• The use of complementary sequence profiles provides rich evolutionary information, contributing to TA-fold's success.
• TA-fold represents a significant advancement in protein structure prediction accuracy.