Figure 8

Correlation between local MQAPs scores and local model quality. Panel A – Pearson's correlation between predicted and observed residue deviation, calculated for CASP7 server models. Hatched bars – correlation for models evaluated by all 3 MQAPs (MetaMQAP, QA_634, and QA_692). Black bars – correlation for all CASP7 server models. In addition we also present the correlation calculated for our CASP7 predictions – QA_038. In CASP7 experiment, QA_038 submitted scores only for a fraction of models (48% of all scorable residues). Panel B – Pearson's correlation between predicted and observed residue deviation as a function of residue difficulty (calculated for CASP7 server models, only for single domain targets). Here we only consider residues scored by all 3 methods (MetaMQAP, QA_634, and QA_692).