Fig. 2

Model performance on the tested proteins, with the color scheme corresponding to the traditionally perceived format complexity (red = complex). (a) Predicted versus true measured aggregate content values at the validation timepoint for all the tested proteins under all formulation conditions. The solid line denotes unity. If the error bars and the solid line intersect, the prediction is deemed correct. Predictions and error bars are schematically described in insert (b)—every point in (a) represents a value predicted by a corresponding model (one example shown), and the error bar corresponds to the 95% prediction confidence interval of the model. The example model in (b) was fitted to data collected up to 6 months (solid markers), the rest is validation (empty markers).