Table 3. Precision and accuracy of the occupancy calculation and the K_d ^cryst value.

	Refined Occupancy (%)		Deduced
	0 mM ASN	100 mM ASN	Kd cryst (mM)
Phenix Occ. + B	63±8	90±6	11
Phenix Occ. only	60±8	98±2	2
Avg. e− count (3 models)	3±2	85±9	17
Model 1: Phenix Occ. & B	4±3	88±9	13
Model 2: Phenix Occ. only	5±5	86±8	16
Model 3: Refmac	1±1	82±11	22

X-ray diffraction data were obtained from 10 thermolysin crystals that were not soaked in asparagine (first column) and from 10 thermolysin crystals that were soaked in 100 mM asparagine overnight (second column). We disregard crystal size because of the long soak times (all crystals were approximately 100 µm). For each group of ten crystals, the average and standard deviation for the refined occupancy are shown separately for each of the methods used for the refinement (two conventional PHENIX refinements, three electron counting methods described in §2.2, and the average of these three). Since the crystals were soaked overnight (t→∞ so that O_max = O_refine) the intra-crystalline dissociation constant K_d ^cryst is readily obtained from O_max using Eq. 2 (shown in the third column) [20]. There is a significant discrepancy between the K_d ^cryst value obtained from the curve fitted O_max (7.5 mM) and the value from the overnight soak O_max (17 mM). The occupancy refinement protocols all have higher precision (as seen by the low standard deviation) than accuracy. This high precision is sufficient to demonstrate that smaller crystals reach high occupancy faster. We report K_d ^cryst to confirm that the binding affinity is within the expected range for a small molecule product, but with significant uncertainty. We did not perform a similar analysis for lysozyme binding to N-acetyl glucosamine because the value obtained from the curve fitted O_max (5.4 mM) was very close to reported values (4–6 mM) [22].