Fig. 6

Distribution of differences in iterative TAP-k scores (for k={1,3,5,20}) between PSI-BLAST and HMMER for the MultiDomainBenchmark Test queries (using the profile generated from searching up to five iterations on a clustered version of the NR database). The average differences (and standard deviations) are 0.16 ±0.25, 0.18 ±0.27, 0.17 ±0.27 and 0.17 ±0.27 for k={1,3,5,20} respectively. A larger area under the curve indicates that PSI-BLAST had more datasets that performed better. Note, the x-axis is sorted independently for each k