Template-based modeling and free modeling by I-TASSER in CASP7

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• Protein structure prediction is crucial for understanding biological function.
• Automated methods aim to achieve accuracy comparable to experimental structures.
• The Critical Assessment of protein Structure Prediction (CASP) experiments benchmark prediction algorithms.

Purpose of the Study:

• To evaluate the performance of the I-TASSER algorithm in the CASP7 protein structure prediction experiment.
• To assess the capabilities of automated protein structure prediction servers.
• To identify limitations and areas for improvement in protein structure modeling.

Main Methods:

• I-TASSER utilizes template-based modeling by threading target sequences through the Protein Data Bank (PDB).
• Global structures are assembled from continuous fragments identified in threading alignments.
• Progressive refinement of structure clusters generates final models.

Main Results:

• I-TASSER achieved high accuracy for template-based modeling targets, with templates improving native structure proximity.
• Correct topology was built for 7 out of 19 free-modeling targets.
• Automated server predictions matched human expert performance across all categories for the first time.
• Model accuracy strongly correlated with template-target similarity (R ≈ 0.9).
• No high-resolution models (<2 Å) were generated for free-modeling targets.

Conclusions:

• I-TASSER demonstrates robustness and potential for genome-wide structure prediction.
• Template similarity remains a key determinant of prediction accuracy.
• Further development is needed for accurate free-modeling target prediction, especially for atomic-level details.