Figure 2

Analysis of individual algorithmic methods and performance of consensuses. All evaluations use 315-chain Balanced Benchmark_2 [8]. A. Evaluation of domain prediction by individual methods using number of predicted domains as a sole criterion. Correct assignments are in green, over-cuts (predicting too many domains) are in red, undercuts (predicting too few domains) are in blue. B. Placement of domain/fragment boundaries by individual methods with respect to secondary structures. Fraction of cuts through alpha-helical structures is indicated in gold, fraction of beta-sheet cuts are indicated in green. C. Fraction of chains that reach consensus. Simple consensus is indicated in blue, weighted consensus indicated in red. D. Fraction of chains whose consensus agrees with expert consensus.