Model building by comparison: a combination of expert knowledge and computer automation

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Structure Prediction

Background:

• The Critical Assessment of techniques for protein Structure Prediction (CASP) trials assess protein structure prediction accuracy.
• Comparative modeling is a key technique for predicting protein structures based on known homologous structures.

Purpose of the Study:

• To evaluate the accuracy of comparative protein model building methods used in CASP2.
• To identify successes and limitations of a specific protein structure prediction approach.

Main Methods:

• Comparative models for four proteins were built using computer algorithms and visual inspection.
• Main-chain modeling involved rigid-body segments from homologues and loop fragments from databases.
• Side-chain conformations were refined using a mean field approach with solvation.

Main Results:

• Accurate sequence alignment between parent and target proteins is crucial and can be misleading if overinterpreted.
• Loop modeling using homologous fragments was effective, but non-redundant fragment libraries posed challenges.
• Side-chain refinement and limited energy minimization improved model quality and stereochemistry.

Conclusions:

• The study identified effective methods for comparative protein model building.
• Areas requiring further attention include improving sequence alignment accuracy and loop modeling strategies.
• The approach demonstrated successes but also highlighted limitations in predicting protein structures accurately.