Fig. 4: Final coexistence is dependent on the sequence of antibiotic treatment.

a An example community of three species (colored circles) for which the sequence of narrow-spectrum antibiotics can affect the final richness. The blue species promotes the action of the antibiotic that targets the orange species, causing the final richness to be non-transitive to the sequence of antibiotic treatment. b (Top left) Remapping of the consumption niche of all species when the death rate of the orange species was increased. The orange species went extinct, as the remapped convex hull does not enclose the supplied resource point. (Top right) Increasing the death rate of the blue species after the extinction of the orange species led to its extinction. (Bottom row) Same as top row, but the sequence of sequential antibiotics was reversed. Now, the green and orange species coexisted after the two treatments, thus \(\triangle \rho=1\). c Population dynamics were simulated until steady state for the scenarios in (b): (top) first with zero death rates, then with nonzero death rate for the orange species, then with nonzero death rate for the blue species (and zero death rate for the now extinct orange species); (bottom) with the reversed sequence of death rates. d An example community with nonzero \(\triangle \rho\) due to the blue species neutralizing the action of the antibiotic that targets the orange species. e Like (b) but for the community in (d). f Like (c) but for the community in (d). g Prevalence of non-transitivity via promotion (red) or neutralization (blue) as a function of the resource competition structure. \(\Delta \rho\) was calculated and the mechanism of non-transitivity was determined for communities of the form in (a) and (d) across parameters (\({D}_{T1}\), \({D}_{T2}\), \({D}_{N}\), \({d}_{1}\), \({d}_{2}\)), for death rates \({d}_{1}\in (0,\,1)\), \({d}_{2}\in (0,\,1)\) and \({D}_{T1}\), \({D}_{T2}\), and \({D}_{N}\) across their entire domains. The fraction of promotion and neutralization were averaged across all combinations \({D}_{N}\), \({d}_{1}\), and \({d}_{2}\). Non-transitivity due to promotion (neutralization) was more likely for low (high) \({D}_{T1}\) and low (high) \({D}_{T2}\). h Same as (g) but averaged across all \({D}_{T1}\), \({D}_{T2}\), and \({d}_{2}\) (or equivalently, \({d}_{1}\) if \({d}_{2}\) was shown on the x-axis).