Figure 3

COGs at varying stringencies. The concept of stringency places a requirement of interconnectedness of elements of a COG. As stringency increases, COGs may split into smaller COGs and less-connected nodes are dropped. Each vertex represents a gene (as used in the initial definition of COGs) or an ORF (as used in this study). Edges represent bidirectional best-hit pairs. Dashed lines enclose elements of a single COG. Grayed vertices and edges do not participate in a COG at the given stringency. There is a single COG of stringency (2) containing all of the vertices in this graph because they are all transitively connected. Stringency (3) COGs are as described by Tatusov et al. [21]. An orthologous group of stringency 3 forms a triangle (such as {i, j, k }); orthologous groups of stringency (3) are clustered if they share two vertices (alternatively: if they share an edge). Stringency (4) OGs are clustered if they share three vertices. The orthologous groups {j, k, l, m } and {l, m, n, o } only share two vertices so they form two separate COGs. At stringency (5) only one orthologous group, and thus only one COG, remains.