Table 1.

The results of predicting real-value distance map and multi-class distance map at the same time versus predicting real-value distance separately on 43 CASP13 hard domains

	L/5 (Precision)	L/2 (Precision)	L (Precision)	MSE	Pearson coefficient
Experiment 1	0.699	0.580	0.446	1.151	0.979
Experiment 2	0.687	0.558	0.430	1.282	0.978

MSE: average mean square error between predicted distances and true distances; Pearson coefficient: the Pearson’s correlation between predicted distance and true distance

Experiment 1: real-value distance prediction by training real-value distance prediction and multi-class distance prediction simultaneously

Experiment 2: real-value distance prediction by training real-value distance prediction alone. The two experiments used the same input features PLM and the same model architecture PLM_Net