CAMEO model quality estimation (CAMEO QE) is intended to continuously evaluate the performance of local model quality assessment programs (MQAPs). The participating groups are, then, allowed to register their method either as Server or as Standalone.
The CAMEO QE targets consist of all the models returned by the public servers participating to the CAMEO 3D category.
Residues in QE targets are classified as "correct" and "incorrect" applying a lDDT-local score threshold of 60 [ref.: PMID: 23986568]. Based on this assignment, the performance of the various local MQAPs are evaluated through ROC and Precision vs Recall analysis being those metrics independent on the exact nature of the different MQAP functional forms. (more)
The assessment is independently made on 4 different datasets of CAMEO QE targets created according to the 'global quality' of the structures expressed as lDDT global score. Namely:
The coordinates of the CAMEO QE target structures are provided in PDB format. Typically in experimental structural biology, the B-Factor field is used to express the uncertainty of the coordinates. Thus, for CAMEO-QE we propose to use the B-Factor field to specify the expected local atomic coordinate of a model. Specifically:
Our baseline predictor is based on sequence conservation. This information is extracted from the PSSM (Position Specific Scoring Matrix) generated by 3 iterations of PSI-BLAST [ref.:PMID: 9254694]. PSI-BLAST is launched using the target sequence as query against and the most recent version of NCBI NR as database.