Evaluation of 3D-Jury on CASP7 models

Area of Science:

• Computational biology
• Structural bioinformatics
• Bioinformatics tools

Background:

• 3D-Jury is an automated consensus method for protein structure prediction.
• It aggregates models from various servers to generate meta-predictions.
• The method is publicly available via the Meta Server (http://meta.bioinfo.pl/).

Purpose of the Study:

• To evaluate the performance of 3D-Jury using models from the Critical Assessment of Techniques for Protein Structure Prediction (CASP7).
• To assess the correlation between the 3D-Jury score and actual model quality (number of correctly predicted residues).
• To examine the utility of the 3D-Jury score as a generic reliability measure compared to original server scores.

Main Methods:

• Analysis of 3D-Jury performance on CASP7 protein structure models.
• Correlation analysis between 3D-Jury scores and the number of correctly predicted residues.
• Comparison of 3D-Jury's best model selection against competing servers.
• Evaluation of 3D-Jury's reliability score against original server reliability scores.

Main Results:

• A significant positive correlation was found between the 3D-Jury score and the number of correctly predicted residues, indicating its predictive power.
• 3D-Jury successfully improved the selection of the best structure model in most cases compared to individual servers.
• The 3D-Jury score demonstrated superior performance in distinguishing good from bad models over original server reliability scores in 27 out of 38 cases.
• A new feature for instant 3D-Jury scoring of user-uploaded models was released on the Meta Server.

Conclusions:

• The 3D-Jury score remains a reliable indicator of protein structural model quality.
• It provides a valuable generic reliability score, particularly for models lacking such assessment from their originating server.
• The Meta Server's instant scoring feature offers a beneficial tool for individual structure modelers to assess their predictions.