Table 2.

Summary of prediction performance. Flex ddG predictions used 50 models and 35000 backrub steps. ddG monomer predictions used the default of averaging the ΔΔG scores of the three lowest scoring output models, as implemented in the original method. ²³ N = number of cases in the dataset or subset. R = Pearson’s R. MAE = Mean Absolute Error. FC = Fraction Correct. Best performance for each metric and dataset is shown in bold.

Mutation Category	Prediction Method	N	R	MAE	FC
Complete dataset	flex ddG	1240	0.63	0.96	0.76
	ddG monomer		0.51	1.57	0.64
	no backrub control		0.56	1.12	0.73
	ZEMu paper		0.61	1.08	0.71
Small-to-large mutation(s)	flex ddG	130	0.65	0.78	0.71
	ddG monomer		0.31	1.55	0.55
	no backrub control		0.41	1.11	0.62
	ZEMu paper		0.48	1.16	0.65
Mutation(s) to alanine	flex ddG	939	0.62	0.96	0.78
	ddG monomer		0.50	1.55	0.66
	no backrub control		0.58	1.06	0.75
	ZEMu paper		0.62	1.03	0.73
Single mutation to alanine	flex ddG	748	0.51	0.75	0.76
	ddG monomer		0.36	1.31	0.62
	no backrub control		0.44	0.90	0.74
	ZEMu paper		0.45	0.86	0.71
Multiple mutations	flex ddG	273	0.62	1.62	0.78
	ddG monomer		0.50	2.44	0.70
	no backrub control		0.58	1.73	0.73
	ZEMu paper		0.64	1.63	0.75
Multiple mutations, all to alanine	flex ddG	191	0.47	1.77	0.84
	ddG monomer		0.34	2.49	0.80
	no backrub control		0.50	1.69	0.81
	ZEMu paper		0.55	1.72	0.79
Multiple mutations, none to alanine	flex ddG	45	0.63	1.38	0.60
	ddG monomer		0.40	2.54	0.38
	no backrub control		0.44	1.66	0.58
	ZEMu paper		0.53	1.59	0.60
Antibodies	flex ddG	355	0.61	0.93	0.74
	ddG monomer		0.50	1.35	0.69
	no backrub control		0.49	1.06	0.72
	ZEMu paper		0.54	1.06	0.67