Table 3.

Performance of various aggregation prediction methods on the SnAggR set.^a

Prediction Methodsb	R2 c	ρc	Area under ROC curved
Structure-based
SAP	0.23	0.48	0.77
DI (Fitted)	0.25	0.50	0.86
CamSol Structure Corrected	0.25	0.49	0.81
AGGRESCAN3D Static	0.00	0.03	0.56
AGGRESCAN3D Dynamic	0.03	0.06	0.55
Sequence-based
CamSol	0.29	0.54	0.82
AGGRESCAN	0.21	0.49	0.80
TANGO	0.35	0.44	0.82
Waltz	0.00	0.01	0.46
PASTA	0.05	0.25	0.68

^a Raw data given in Table S5.

^b See Methods section for details.

^c Squared Pearson correlation coefficients and Spearman rank-order correlation coefficients are based on dimer LogK_d(μM) data determined by analytical ultracentrifugation and listed in Table 2 and Table S5.

^d Area under ROC curve (see also Figure 7(b)) values are based on dimer % data determined by analytical ultracentrifugation and listed in Table 2 and Table S5.