Fig. 6.

Illustration of why CMsearch is able to detect better homologs and rank them at the top for the three cases in Figure 5: each triangle represents a sequence; each pentagon represents the corresponding structure; each solid line represents a strong link; each dashed line represents a (relatively) weak link; the top three homologs predicted by HHsearch (in blue) have strong query-sequence links; the top three homologs predicted by CMsearch (in red) have weak query-sequence links but strong links among the sequences and structures of them three; and this demonstrates that CMsearch is capable of compensating weak links to true homologs that are consistent (i.e. the true homologs tend to be similar to each other)