Table 2:

CI in the binary and quantitative data sets^a

Setting	E	IC	GPCR	NR	Kd Q	Kd B	Ki Q	Ki B
S1	96.0	96.4	92.7	86.1	88.3	95.2	79.3	93.4
S2	83.7	80.2	85.2	84.6	74.8	77.5	73.6	85.5
S3	92.1	94.0	89.4	73.8	86.1	93.6	66.6	85.0
S4	76.4	67.8	78.6	67.7	67.0	70.0	59.2	74.9

^aThese summary results were based on the normalized SW sequence similarity and 2D structural similarity. Data sets E, IC, GPCR and NR were originally in binary interaction format, whereas the kinase K_d and K_i were originally quantitative data. The full set of prediction results using different prediction models, cross-validation approaches and evaluation metrics are provided as Supplementary Tables S1–S17. All prediction accuracies reported in this work differ significantly from random (P < 0.01, permutation test). Q, quantitative data; B, binarized data.