With this method, the residue error on edges can be removed separately and has potential to reduce total process time.
The selected models were evaluated by the Z-scores, the discrete optimized protein energy values, and the residue error plot (SwissProt website, http://swissmodel.expasy.org/workspace/index.php?func=tools_structureassessment1, last accessed August 2014).
The subscription im stands for the FIR function of neuron i on muscle m, and ε stands for the residue error, and N is the number of total neuron units involved into the prediction.
QMEAN [ 38] at http://swissmodel.expasy.org/qmean/cgi/index.cgi page=help is a composite scoring function which is able to derive both global (i.e., for the entire structure) and local (i.e., per residue) error estimates on the basis of one single model.
Supplementary Figures S11 (boxplot) and S12 (ROC analysis) show the performance of QMEANlocal in estimating per-residue errors on all CASP8 models.
We demonstrate the value of the QMEAN Z-score as a statistically well-founded measure of absolute quality and end with a critical discussion of the limitations of this approach for predicting absolute local per-residue errors.
It predicts per-residue error in multiple fold recognition models which is followed by an analysis of the conservation of the per-residue errors across all models (McGuffin, 2008).
Based on these analyses, we further examined the per-residue error (Fig. S5), visualized using a color gradient from blue (reliable regions with estimated error below 1 Å) to red (potentially unreliable regions with estimated error above 3.5 Å).
The prediction of local (per-residue) error estimates is an active field of research.
We confirmed that, for almost all residues, the error rate distributions among the 50 control samples fitted a Poisson distribution (data not shown).
An improvement of similar magnitude is seen after allowing for alignment errors (residue shifts).
