Figure 4

ROC for finding a domain in the full protein structure. Full red line: the method discussed in the text, with δ = 0.5 (15 CPU min). Dotted orange line: SSM [12](several tens of CPU), dash-dot, green: SGM ([24], Full green line: "generation 2000" high resolution methods, in the same order of appearance as in Fig. 3 (by descending size of the area under the ROC curve): 3dhit [9](6 CPU hrs), TMalign [13](8 CPU hrs), SABERTOOTH [14](5.5 CPU hrs), MAMMOTH [8](3.5 CPU hrs). These methods, as well as SSM, do the high resolution alignment, which both makes them more reliable in the high specificity (small FP) region of the graph, and slows them down.