Fig. 2

Optimal experimental designs. This figure shows the optimal composition of the seeder and contact populations for different experimental designs: a Single contact group design: ~ 15% of individuals are seeders, where seeders have genotype BB (or AA) and contacts predominately have genotype AA (or BB), with ~ 10% BB, to allow for estimation of the susceptibility SNP effect \({a}_{g}\). Estimation of dominance was found to be challenging using only a single contact group (not shown). b Multiple groups “pure” design: ~ 47% of individuals are seeders. Seeders and contacts consist of different combinations of AA and BB across groups (and AB when dominance is investigated). c Multiple groups “mixed” design: a small number of individuals are seeders (typically two or three, sufficient to initiate epidemics). When dominance is not investigated, there is a 83%/17% split in AA/BB individuals in the contact population in group 1 and vice-versa in group 2. When dominance is investigated, there is a 80%/10%/10% split in AA/AB/BB individuals in the contact population in group 1, and these proportions are permuted to define the two other groups. Optimisation of these designs was (for the most part) based on maximising the precision with which the infectivity SNP effect \({a}_{f}\) can be estimated (since this was generally the most difficult trait to estimate). However in cases where maximal precision for \({a}_{f}\) corresponds to minimal precision for \({a}_{g}\), values are chosen to give equal precision to the two (e.g. ~ 10% BB in a, as discussed in the paper). The percentages above are, to a large extent, independent of \({R}_{0}\) (see Additional file 8) or other factors in the model/data (see Additional file 11). For reference the optimal homozygote balance \({\chi }_{\mathrm{seed}}\) and \({\chi }_{\mathrm{cont}}\) (i.e. proportion of AA minus BB individuals) and homozygosity \({H}_{\mathrm{seed}}\) and \({H}_{\mathrm{cont}}\) (i.e. proportion of AA plus BB individuals) are shown for each design (the ‘\(\approx\)’ symbols indicate that these are optimal values to be aimed for, accounting for the fact that the number of individuals is discrete). The same basic designs can be replicated multiple times within an experiment. Note that the results equally apply to the estimation of non-genetic factors, e.g. vaccination effects (AA replaced with “Vac.” and BB replaced with “Unvac.” and dominance not applicable). The spatial separation between seeders and contacts in this diagram is for illustrative purposes only