Table 2. Statistics from AutoSol runs of various U-derivative data sets using the single-wavelength anomalous diffraction (SAD) method.

AutoSol was rerun under identical conditions for test cases in this study assuming that there were eight HA sites. The results of the U123 SAD AutoSol run with the partially interpretable initial experimental maps are shown in bold. Interestingly, its inverted HA structure also produced a clear solvent boundary as apparent from the relatively low R-factor statistics between the observed and back-transformed amplitudes during density modification. FOM stands for figure of merit for experimental phases. When correct solutions are found, one expects a large difference in overall score between the correct (original) and inverted HA substructure solutions. In addition to a low R factor in density modification, a large map skew value and high correlation coefficients (CC) are good indicators of a correct solution at medium and high resolution. They may not be as reliable at low and medium resolutions. Thus, visual inspection of resulting experimental maps remains the best method to evaluate the correctness of the HA substructure solutions.

	Heavy-atom search statistics				Density-modification statistics
U HA hand	Anomalous resolution (Å)	Sites	FOM	Overall score	R factor (%)	Map skew	Density CC
U1 original	5.10	10	0.334	31.79 ± 16.85	39.32	0.11	0.73
U1 inverted				30.89 ± 17.02	39.28	0.11	0.73
U2 original	4.45	6	0.336	21.94 ± 17.04	45.15	0.06	0.81
U2 inverted				21.82 ± 17.02	45.52	0.06	0.82
U3 original	5.74	6	0.321	23.10 ± 17.16	40.20	0.07	0.71
U3 inverted				21.10 ± 16.90	39.50	0.05	0.68
U12 original	5.67	7	0.287	21.59 ± 16.97	39.06	0.06	0.65
U12 inverted				19.43 ± 16.49	Failed
U23 original	3.75	10	0.304	22.70 ± 17.13	39.77	0.06	0.72
U23 inverted				21.99 ± 17.04	41.23	0.06	0.71
U123 original	6.07	7	0.376	34.25 ± 16.14	25.70	0.12	0.76
U123 inverted				24.39 ± 17.20	28.11	0.07	0.65