Figure 1.
Illustration of the iterated high throughput experimental process.
Note the predictive modeling procedure in the loop.
Figure 2.
Rank order of all tested formulations found with Evo-DoE vs. the standard recipe.
Error bars on the best new recipe were taken from three repeats performed on the same day (space repeats, see Materials and Methods). The error bars from the standard were taken from 75 total repeats performed over the course of the experiment.
Figure 3.
Representation of a 3-dimensional section of experimental space.
For each PG-type lipid, shown in bold, the horizontal section lists the lipids from the group with a net negative charge and the vertical section lists the reagents in the aqueous phase and their corresponding pH values. Response levels (the UV/Vis absorbance of Amphotericin B associated with the formulation): dark grey, >0.20; medium grey, 0.15–0.20; light grey, 0.10–0.15; white, <0.10; Blank cells, not determined. For abbreviations, see Materials and Methods.
Figure 4.
Change over time in lipid components in recipes tested.
The representation of each lipid, computed as a percentage of recipes that have each particular component, for each generation of Evo-DoE is shown for seven successive generations. For generation 2–7, only model-based recipes are considered. A) The PG lipid group; B) the negatively charged lipid group. See Materials and Methods for abbreviations and groupings.
Table 1.
Selected high-response formulations.
Table 2.
Aqueous phase library.
Table 3.
Lipid library.