Fig 1.
The effects of spores and toxin on mortality in Bacillus thuringiensis.
A) The mortality rates for varying quantities of spores of Bacillus thuringiensis a Cry null strain supplemented with 60 pg of toxin (open circles, dashed line) or 180 pg of toxin (solid circles, solid line). A glm of the form y ~ Toxins + log(Spores+1) + Toxins* log(Spores+1) is fit and shown for both toxin doses (see S2 Table in S1 Text for parameter details) with S.E. shown. In both cases spore dose has a positive impact on mortality, and in the 60 pg case, there is a gap between zero spores and the next lowest dose. B) The mortality rates for 900 Bacillus thuringiensis spores supplemented with varying quantities of toxins (+/- S.E.) in two experiments. The fit curves are from a nonlinear regression with an exponential model (Eq 1). This IAH model is not sufficiently threshold-like to describe the data; it overestimates low dose mortality.
Fig 2.
The dose-response for wild type Bacillus thuringiensis (+/- S.E.) in two experiments.
The best IAH model (beta-Poisson) fit from all doses over 20% mortality is also shown. Again dashed line corresponds to the first experiment (with open circles), and solid line corresponds to the second experiment (with solid circles).
Fig 3.
The distribution of the differences between predicted mortality and ‘actual’ mortality from best-fit models (fit to all doses over 20% mortality) at low doses in simulated experiments (generated from bootstrapping data from each dose 5,000 times).
The difference between predicted and “actual” low-dose risk when the model is chosen between the best fit exponential and beta-Poisson models based on AIC for experiment 1 focusing on the spore dose of mean 150 (A) and experiment 2 focusing on the spore dose of mean 130.59 (B). The difference between predicted and “actual” low-dose risk when the less commonly used, exact (confluent hypergeometric) form is shown for data in experiment 1 (C) and experiment 2 (D).