To split or not to split: CASP15 targets and their processing into tertiary structure evaluation units.

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• The Critical Assessment of protein Structure Prediction (CASP) experiment assesses computational methods for predicting protein 3D structures.
• Accurate protein structure prediction is crucial for understanding biological function and disease.

Purpose of the Study:

• To process CASP15 protein targets into standardized evaluation units (EUs).
• To assign these EUs into prediction classes based on difficulty and methodology.

Main Methods:

• Protein targets were segmented into structural domains based on compactness and similarity.
• Domains were combined into EUs if prediction performance remained consistent, or retained individually if performance improved.
• EUs were categorized into template-based modeling (TBM), free modeling (FM), or overlap classes.

Main Results:

• 112 EUs were generated from 77 tertiary structure prediction targets.
• EUs were classified into four categories: TBM (easy/hard), FM, and TBM/FM overlap.
• Over one-third of CASP15 EUs were classified as FM, indicating significant prediction challenges.

Conclusions:

• The developed EU processing and classification system provides a structured approach for evaluating protein structure prediction methods.
• The high proportion of FM targets highlights the ongoing difficulty in predicting protein structures without detectable homology.
• This classification aids in understanding the performance landscape of different prediction strategies in CASP assessments.