Figure 1
From: Comparative assessment of performance and genome dependence among phylogenetic profiling methods
Method performance comparison between SG, DM, EM and SM. On the y-axis is the percent of predicted associations sharing at least one KEGG pathway, when looking at the N highest confidence predictions (x-axis). Four methods are compared: SM, a binary profile generated from BLAST E-values using 10-5 as a cutoff for presence/absence; DM, a multinary profile that discretizes BLAST E-values into 11 equidistant bins; SG: a multinary profile that uses three bins in E. coli and 5 bins in S. cerevisiae, where presence/absence boundaries are optimally determined, and intermediate binning boundaries are equiprobable; EM, a profile consisting of normalized BLAST bit-scores. Profile similarity was measured for the SM, DM and SG methods using mutual information, while Pearson's correlation was used for the EM method. (A) E. coli and (B) S. cerevisiae are used as the query organisms, with SG in red, DM in green, SM in black and EM in blue. The profile genome composition used in both graphs is the Prok180 set, which is composed of entirely prokaryotic genomes.