TABLE 2

The probability (%) of detecting balancing selection

Width of window
Infinite sitesFinite sites (1000 bp)Finite sites (333 bp)
ρ0.0250.050.10.210.0250.050.10.210.0250.050.10.21
188.191.493.394.388.285.690.292.893.583.875.083.889.491.576.5
384.788.288.987.157.680.885.386.383.544.769.276.679.276.432.2
1076.777.373.662.07.769.069.964.551.04.155.257.652.339.92.4
3059.353.240.722.00.248.142.531.015.70.137.234.323.811.50.0
10026.020.210.93.10.020.515.68.92.40.015.412.36.92.10.0
10a3.76.16.85.20.33.76.06.85.10.33.25.66.85.10.2
  • Power was estimated from 5000 replicates (1000 for ρ = 100) using θ = 10 and n = 24 (12 in each allelic class). Selection was deemed to have been detected if D > 2.01 (Tajima 1989) in a window of the specified size. In the finite-sites models, infinite-sites mutations (i.e., random numbers in the unit interval) that were sufficiently close to each other were deemed to have affected the same site and simply reverted it to its previous state.

  • a Loss-of-function model, with rates of loss and reversion equal to 1 and 0.01, respectively.