Fig. 8.27. The tree structure of the top rule of the Dayhoff GP model.

op rule has explored one interplay between ܴଷ (labelled by c) and

led by d) with specific amino acids Y and W, which may have a

pact on the discrimination power between the cleaved and non-

actor Xa protease peptides. Figure 8.28 shows the histogram of

ctions using the top rule shown in Figure 8.27. It can be seen that

ictions demonstrated a good bimodal distribution for the

ation between two types of peptides. The next top rule was

YdReU(aIdK)&)+, whose fitness measurement was 0.84 and

UC was also 0.93. This rule revealed the interplay between ܴଵ

by a) and ܴସ (labelled by d).

he histogram of the predictions using the top rule generated by the Dayhoff GP

using the BLOSUM62 mutation matrix, the GP learning was

d at the 94^th cycle. The fitness measurement and the AUC of the

were 0.87 and 0.96, respectively. Figure 8.29 shows the evolving

the fitness measurements and the AUC values.