Table 3.

A comparison of predictive performances obtained by using random sampling versus balanced sampling for training reveals that both protein representational biases and sequence features contribute to PPI predictions

The subset sampling for training	Yeast		Human
	Random	Balanced	Random	Balanced
M1	0.71 ± 0.02	0.65 ± 0.01	0.72 ± 0.01	0.63 ± 0.01
M2	0.67 ± 0.02	0.60 ± 0.02	0.67 ± 0.01	0.57 ± 0.01
M3	0.57 ± 0.01	0.55 ± 0.01	0.58 ± 0.01	0.54 ± 0.01