The GARP refinement procedure. Schematic example for the calculation of a mutual clustering coefficient (C
) and window averaging. Residues (nodes) are shown as circles, predicted interactions (edges) are shown as black lines, and part of the protein backbone sequence is shown as thick grey lines. The analyzed edge (v, w) is shown in bold with some other edges of the graph. Edges, which connect nodes v and w through a third node, are shown as regular lines. The mutual clustering coefficient compares the number of these edges, or their integrated weights, with the number, or weights, expected from all edges, in which v or w are present. The five top and five bottom nodes form the sequence window for nodes w or v, respectively. Edge (v, w) is given the mean C
value of all edges between the two windows.