CASP13 target classification into tertiary structure prediction categories.

Area of Science:

• Structural biology
• Computational biology
• Bioinformatics

Background:

• Protein structure prediction is crucial for understanding biological function.
• Critical Assessment of Structure Prediction (CASP) provides a benchmark for evaluating prediction methods.
• CASP13 involved assessing protein target structures using various computational approaches.

Purpose of the Study:

• To define evaluation units (EUs) for protein targets in CASP13.
• To categorize EUs for assessing template-based modeling (TBM) and free modeling (FM).
• To establish criteria for classifying target difficulty and prediction performance.

Main Methods:

• Splitting 80 protein targets into 112 EUs based on structural domains and template organization.
• Classifying EUs into TBM or FM categories using target difficulty.
• Utilizing the Evolutionary Classification of Protein Domains (ECOD) database for evolutionary relationships.
• Applying objective measures for sequence, structure, and performance clustering.
• Analyzing target difficulty versus server performance using scatter plots.

Main Results:

• EUs were successfully categorized for assessing different modeling approaches.
• Targets with close template homologs and good server predictions were classified as TBM.
• Targets with poor server predictions, including fast-evolving homologs, analogs, and new folds, were classified as FM or FM/TBM overlap.

Conclusions:

• The EU splitting and categorization strategy effectively addressed target difficulty in CASP13.
• This approach facilitates accurate assessment of protein structure prediction methods.
• The classification scheme aids in distinguishing between template-based and free modeling challenges.