Enhancing fragment-based protein structure prediction by customising fragment cardinality according to local secondary structure

Table 3 Performance comparison between generating 20,000 and 2000 decoys in both standard and SS-based predictions

	First model	Best model	Average of the Best 5 models
Std. Predictions 2000 vs 20,000 Decoys	17/24 (71%) + 26.3%	19/24 (79%) + 7.6%	19/24 (79%) + 8.2%
SS-based Predictions 2000 vs 20,000 Decoys	18/24 (75%) + 30.2%	19/24 (79%) + 11.7%	19/24 (79%) + 14.7%

ISSN: 1471-2105