The ResiRole server to enable assessments of structure prediction techniques using functional site predictions
Joshua Toth, Paul DePietro, Juergen Haas, William A. McLaughlin · 2020
The Continuous Automated Model Evaluation (CAMEO) platform, https://www.cameo3d.org/ , presents the results of protein structure predictions generated by the hosted structure prediction servers for prerelease sequences from the Protein Data Bank, https://www.rcsb.org/ . Here we describe the ResiRole server, http://protein.som.geisinger.edu/ResiRole/ , for the assessment of structure models available through CAMEO regarding their abilities to have functional site predictions like those at corresponding sites in their experimentally determined reference structures. The results are presented as average difference scores per structure prediction technique and per structure model, where each difference score is defined as the absolute difference in the cumulative probability of the functional site prediction in the reference structure and that at the corresponding site in the structure model. Results are accessible according to target difficulty based on lDDT score ranges. The difference score is compared to other metrics for estimating structure model quality which use the reference structures for bases and is found to be a complementary and informative quality metric. For example, when using difference scores as the benchmark, we find that the quality of structure models produced by NaiveBLAST is, on average, underestimated. The results indicate that NaiveBLAST models may contain more information about local functional site predictions than previously estimated. Support for this project was partly provided by NIGMS [grant number 5U01 GM093324-02].