Skip to main content

Table 1 Qualitative assessment of our comparative modelling methods at CASP experiments.

From: A comprehensive analysis of 40 blind protein structure predictions

Category

CASP1

CASP2

CASP3

CASP4

Alignment quality

poor

fair

fair

fair

Side chains

~50%

~75%

~75%

~75%

Short loops (≤ 6 aa)

~3.0 Ã…

~1.0 Ã…

~1.0 Ã…

~1.0 Ã…

Longer loops (> 6 aa)

> 5.0 Ã…

~3.0 Ã…

~2.5 Ã…

~2.0 Ã…

  1. The results given are for predictions for which there was little or no alignment error (since we relied on publicly available webservers for alignments). For evaluating side chain predictions, the percentage of χ1 torsion angles predicted within 30° on average is given. For evaluating variable main chain (loop) predictions, the average of the Cα root mean square deviation (RMSDs) (calculated using a global superposition of the target and the model) is shown. The major improvement in our methods from CASP1 to CASP2 is from the use of manually-curated alignments and the development of a graph-theory approach to handle the interconnectedness problem in protein structures [35].