Fig. 2

Results of simulations based on the model proposed by Lighten et al. Parasite haplotypes (yellow) and host supertypes (black and other colours) are represented as coordinates in 1000 × 1000 grid, reflecting their functional properties (the closer a parasite is to the host, the more likely is a successful host immune response). Each panel shows how changes in specified parameters in the model, compared to the parameters used by Lighten et al. to produce their Fig. 4, affect the outcome (parameters other than stated in the description are as in Lighten et al.). a Mutation rate set to 10⁻² for pathogens and 10⁻³ for hosts, population size increased to 100,000. In contrast to the results reported by Lighten et al. using much higher mutation rates, the effective number of alleles maintained in a population is small. b Selection on parasites (but not on hosts) ‘turned off’. This simulation gives the most similar outcome to that reported by Lighten et al., despite the lack of host-parasite coevolution (see Supplementary Fig. 6 for the scenario with no random pathogen genotypes added). c Simulations that do not seed each pathogen generation with ≈100 new genotypes, but mutation parameters as in Lighten et al. Host-parasite coevolution utilizing mutational variance alone does not maintain several stable supertypes, even though the simulations started from creating a set of random MHC alleles and pathogens in the same way as Lighten et al. d Parameters as in c, but 10 independent parasites simulated. Several supertypes are observed at any time point, but they are not stable through time. Effective number of alleles (#alleles) has been calculated for a sample of 100 individuals to allow comparison with the results of Lighten et al. The simulations were written in MatLab and the algorithm follows that described by Lighten et al. The MatLab code is provided in our Supplementary Data 1