TABLE 3

Neutrality vs. selection in nonequilibrium populations (deep bottlenecks)

Model ParameterPerformanceSFSF*ωω*SVM
Sweep in phase 1TP (FP = 0.51)0.640.660.390.490.71
Median distance (bp) from target (SD)10813 (6768)10497 (6832)11986 (6595)10239 (6186)
Random target distance (SD)11053 (6827)11308 (6803)11575 (6645)11944 (6945)
Sweep in phase 2TP (FP = 0.20)0.620.640.360.440.73
Median distance (bp) from target (SD)9666 (6531)10828 (6896)11854 (6500)10469 (6123)
Random target distance (SD)11508 (6885)11397 (6808)11877 (6750)11555 (6804)
Sweep in phase 2*TP (FP = 0.08)0.720.780.630.120.97
Median distance (bp) from target (SD)9512 (6659)10986 (6977)10905 (6482)11328 (6487)
Random target distance (SD)12067 (6983)12265 (6920)11647 (6950)13236 (7213)
Sweep in phase 3TP (FP = 0.56)0.530.550.480.460.63
Median distance (bp) from target (SD)10377 (6831)10845 (6833)11342 (6662)10624 (6541)
Random target distance (SD)12202 (6908)11641 (6860)12151 (6920)12220 (6824)
  • A deep bottleneck, named model A, is examined. The ratioEmbedded Imageand the length of the bottleneck is 0.00075. A beneficial mutation may appear within each phase of this three-epoch model (where time is measured backward in units of 4N generations): a recent sweep at time 0.01 (sweep in phase 1), a sweep within the bottleneck at time 0.0107 (sweep in phase 2), and an old sweep at 0.115 (sweep in phase 3). The selection coefficient is 0.002. Additionally, in the “sweep in phase 2*” model we describe a sweep that completes within the bottleneck (s = 0.8). The true positive rates of the neutrality tests are shown for each sweep model. The other rows depict the distance between the predicted and true targets and the random expectations for the distance.