Linearity of QI-CU relationship

CU values were measured on clustering results generated under different edge score cutoff values. Four data sets of one-to-one COG, EC and PFam family correspondence were used. CU is seen to be linearly related with the QI, the clustering quality index in Figure 1. R value has a minimum of 0 and a maximum of 1. Values close to 1 indicates the relationship predicted by an employed curve is very good. The minimum was 0.89, and the maximum was as high as 0.99. This implies clustering by maximizing the global CU value can produce the optimal clustering result sought.

Legend            X-axis: QI        y-axis: CU

(a) Set1                                         (b) Set 2

                        (a) Set 3                                        (b) Set 4

Figure 1. The linear relationship between QI and CU