Table 1.

The comparative results of the prediction performance of the method based on different negative data sets, respectively, using AC with lg of 25 amino acids

Negative data set	Psub	Prcp	1-let	2-let	3-let
Sensitivity (%)	85.22	41.76	79.29	69.81	60.74
Precision (%)	87.83	62.64	82.67	85.14	80.15
Accuracy (%)	86.23 ± 1.95	58.42 ± 1.68	79.25 ± 7.80	77.30 ± 12.38	70.25 ± 10.40

Psub is the negative data set of non-interacting pairs of non-co-localized proteins; Prcp is the negative data set derived from the method by Shen et al. (26). The three negative data sets, 1-let, 2-let and 3-let are obtained by shuffling the protein sequences with k-let counts, k = 1, 2, 3.