The best performing formulations (highest object counts) were identified from each screen and taken forward as the basis of the design of the more complex formulation space to be evaluated in the next stage. A linear strategy inherently risks missing any dramatic synergistic effects between excipients that are never tested in combination (having been eliminated see more from consideration during earlier steps) and

the true maxima in concentration space (which is only explored coarsely). To reduce these risks, 4 additional screens aimed to cover both a broader sampling of the overall formulation space (‘shotgun’ screens) or to finely explore concentration effects of promising formulations (‘targeted’ screens) were interspersed in the process. A total of 11,823 unique formulations (as defined by combination of excipients, excipient concentrations, and pH) were screened in 35 HT screens comprising 5 stages of linear screening and additional non-linear screens (Table 1, full and summarized datasets in Supplementary Data Online). Intra-assay variability was typically in the range of 10–25% RSDs normalized across control formulations, and all assays reported had RSDs below 30%. The highest performing formulations (based on rank ordered normalized object counts) were selected at each stage as the basis of the design

of the subsequent stage. Pairwise comparisons of formulation performance quoted are significant at the p < 0.05 level by standard t-test, with 4–10 replicates per ABT-888 solubility dmso formulation. A small number of datapoints attributed automation error were removed from the calculations. In general, as the complexity of the formulations increased

with progression through the stages, the performance of the top formulations from each stage increased. Increases in performance were incremental or additive L-NAME HCl at best, and no truly synergistic effects (AB ≫ A + B) were observed. Stage I was designed to broadly assess the effect of buffers on viral stability (29 variables, 218 unique formulations). Citrate pH 7.4, citrate pH 6.0, potassium phosphate pH 7.4, and histidine pH 7.4 were identified as the highest performing buffers. In Stage II, they were combined with stabilizers (73 variables, 3134 unique formulations). Formulations containing gelatin, valine, citrate, and trehalose were typically high performing, and citrate pH 6.0 was generally the best performing buffer background. In Stage III (50 variables, 2740 unique formulations), higher order combinations of the same excipients used in Stage II yielded increased performance. A non-linear screen examined the effects of varying the concentrations in two high-performing quaternary formulations identified in Stage III (Fig. 3a).