However, a more extreme drift model is plausible, since this role for selection assumes that all modifications are under strong selection, a contention which has not been demonstrated.
This is because symmetric balancing selection assumes that large numbers of alleles are equivalent to each other in their fitness, and thus the probability for producing descendants is also equal.
The simplest population-genetic model of natural selection assumes one autosomal locus with two alleles, A1 and A2, as above.
However, theory that underlies standard scans for selection assumes a simple mutation model.
Model M3 (selection) assumes three discrete site classes (ω0, ω1, and ω2) with three corresponding proportions (p0, p1, p2) estimated from the data.
Model A (branch-site test of positive selection) assumes four classes of sites, the first two classes having omega values 0 < Δ0 < 1 and Δ1 = 1, assuming purifying selection and neutrality, respectively.
Namely, the methods of population genetics used to detect selection assume that positively selected variants emerge according to the infinite sites model (ISM; Kimura 1969).
Figure 1 illustrates this point by comparing greedy seed selection (assuming perfect link information) and mean random-seeding performance across all possible budgets in a 100-node network.
This is because most methods used to infer positive selection assume that the phylogenetic tree and branch lengths are constant across all sites in the alignment, a tenet that is invalid in the presence of recombination.
We thus selected a background threshold (bgThreshold = 100; see Parameter Selection), assuming that counts below it do not accurately reflect strain abundance.
