Skip to main content
Figure 6 | BMC Bioinformatics

Figure 6

From: On the necessity of dissecting sequence similarity scores into segment-specific contributions for inferring protein homology, function prediction and annotation

Figure 6

HMMER2 versus HMMER3 average domain score (averaged over all). When the fold-critical and remnant scores (see Table 1, columns 5 and 6) were considered, the distinction between the true and false hits becomes apparent. The Y = X margin depicts two regions: above is where the fold-critical E-values were smaller than the reminant E-values and below as vice-versa. As a general trend, the fold-related scores of hits IF2P_HUMAN, IF2P_MOUSE, IF2P_PONAB, NUCL1_ORYSJ, MLL2_MOUSE, CORTO_DROME, DHKL_DICDI, AMOT_MOUSE, NUCL_HUMAN and PK4_DICDI (see red points) were much smaller than the remnant scores indicating that they are spurious hits and their corresponding fold-related E-values spans from 2.9e-2 to 2.6 (HMMER2) and 3.6e + 1 to 2.0e + 5 (HMMER3) against the more significant remnant segments’ E-values ranges of 3.6e-6 to 1.0e-4 (HMMER2) and 5.5e-27 to 3.3e-1 (HMMER3). In contrast, the fold-related scores were larger than the remnant scores for hits CHLI_PORPU, HEM1_METKA (see blue points). For TIP12_MAIZE (see blue point), the difference between its fold-related and remnant scores was marginal. The corresponding fold-related E-values of 1.2e-4 to 2.2e-6 (HMMER2) and 3.1e-3 to 1.5e-9 (HMMER3) were more significant than the remnant segments’ E-values of 9.4e-5 to 1.4e + 2 (HMMER2) and 3.1e + 2 to 5.3e + 5 (HMMER3).

Back to article page