Evidence theoretic protein fold classification based on the concept of hyperfold

Area of Science:

• Computational biology
• Bioinformatics
• Structural biology

Background:

• Assigning protein domains to fold classes is complex and controversial.
• Identifying protein domain fold patterns solely from sequence data presents significant challenges.

Purpose of the Study:

• Introduce novel concepts, hyperfolds and interlaced folds, to address uncertainty in protein fold classification.
• Develop a framework for handling ambiguity in predicting protein domain fold patterns from sequence information.

Main Methods:

• Utilize sequence-based features for predicting hyperfolds.
• Employ the Dempster-Shafer theory of evidence, including bodies of evidence and Dempster's rule of combination.
• Develop a classification architecture to assign query proteins to hyperfolds.

Main Results:

• The developed classification architecture was applied to identify protein folds within 27 SCOP fold patterns.
• The approach demonstrated potential for improved results compared to existing predictors on a benchmark dataset.

Conclusions:

• The novel hyperfold concept and Dempster-Shafer based framework offer a promising approach to protein fold classification.
• This method effectively handles the inherent uncertainty in predicting protein domain fold patterns from sequence data.