Figure 4

Performance of existing algorithms and our method for the prediction of protein complexes. Each graph is a precision-recall graph in which the vertical axis indicates the precision and the horizontal axis indicates the recall. Red lines (denoted as "DDI verified") show the results of our method and blue dotted lines indicate the results of existing algorithms (either MCL, MCODE or clustering coefficient, denoted as "Original"). Each algorithm was executed with three DDI data sets, DDI dataset (A), DDI dataset (A + B) and DDI dataset (A + C). The inflation parameter, which is only a configurable parameter in MCL, was changed from a minimum value to 1.2 to 4.8 in steps of 0.6, and MCL was also executed at a maximum value 5.0. The Node Score Cut-off parameter, which is the most influential parameter in MCODE, was changed from 0.0 to 1.0 in steps of 0.1. The threshold in the clustering coefficient was changed from 0.1 to 1.0 in steps of 0.1.