Table 3.

Result of validation of the effectiveness

Methods	Data1_4R TPR	Data2_6R TPR
Both	0.524	0.361
sequence only	0.333	0.361
structure only	0.524	0.250

We compared binding-affinity prediction performances in three types of analyses; first one is the analysis based on sequence similarity only, second one on secondary structure similarity only, and the last one on both parameters. The evaluation metric was is ‘True Positive Rate (TPR) when False Positive Rate (FPR) is 0’. We used data from rounds 4 and 6, the last round data from Data1 and Data2, respectively. For each dataset, the highest values are shown in bold.