Conclusions and Discussion - Genetic Diversity in Establishing Plant Populations: Founder Number and Geometry - page 332

Table 18.3 Demonstration that, of two trials differing only in the geometric placement of
founders, the trial that loses lower-frequency unique alleles at faster rates can exhibit greater
standard deviations in heterozygosity when loci with higher-frequency alleles only are
used.

Percent change in expected heterozygosity standard deviation from the

founding generation to:

Generation 1 Generation 3

Lower Higher Lower Higher

unique allele unique allele unique allele unique allele

loss trial “o” loss trial “r” loss trial “o” loss trial “r”
Trial allele panels: conditions conditions conditions conditions
A) 0.5, 0.5 57.1 69.2 (21.2) 107.1 123.1 (14.9)
B) 0.17, 0.17, 0.17, 0.17, 0.16, 0.16 60.0 60.0 (0) 110.0 130.0 (18.2)
C) 0.34, 0.33, 0.33 38.5 57.1 (48.3) 92.3 107.1 (16.0)
D) 0.78, 0.11, 0.11 22.0 26.3 (19.5) 39.0 47.4 (21.5)
E) 0.9, 0.043, 0.33, 0.33 21.9 24.0 (9.6) 37.0 43.7 (18.1)
F) 0.6, 0.3, 0.07, 0.03 20.4 29.4 (44.1) 38.9 51.0 (31.1)
These results suggest that the trial conditions that lose more low-frequency unique alleles will
have greater variance in the frequencies of more common alleles across replicate populations
developing from the same starting conditions. Here, two trials, type o versus r, were identical
except for a decrease in space between founders in trial r, which induced a higher rate of loss of
low-frequency unique alleles (see Figs. 9.9 and 9.10, and the associated discussion in Chapter 9).
Trials type o versus r were re-run in a series of sets of comparative trials. In each comparative
trial set (o versus r), only one locus was used that had a unique panel of alleles (see trials
A through F in the table) that were at higher frequencies than used throughout most of this
book. For example, in trial set A, both the o and r trial initial conditions were held constant to
those used to generate Figs. 9.9 and 9.10, but here only one locus was used, the locus having
two alleles of equal frequency. The data reported are values taken from 1,000 replicate runs
per trial for trials A through C, and 10,000 replicate runs for trials D through F. The values
shown are the percent change in standard deviation of the expected heterozygosity from that
in the founding generation to generation 1 (after one round of reproduction) or generation 3.
Values in parentheses give the percent difference between trials o and r run under equivalent
starting conditions except for founder geometry. Standard deviation increased in all type r
trials (which lost low-frequency unique alleles at higher rates: Fig. 9.10) to a greater degree
than the corresponding type o trials in all cases but one (in which there was no difference).
These results suggest, comparing two sets of founding conditions that are identical except
for founder geometry, that the set producing the highest rate of loss of low-frequency unique
alleles can exhibit greater standard deviation for expected heterozygosity. This result is likely
due in part to greater variance in losses of copies of higher-frequency alleles yielding greater
variance in the frequencies of such alleles through generations across repetitive trials. Further,
it implies that trial conditions in which more alleles are lost will produce a greater range of
random changes of the probabilities of occurrence of particular multilocus allele combinations
across repeated identical trials or founding events.

The results described in Table 18.3 show that when two sets of founding
conditions are identical, except for the geometric placement of founders
such that one trial loses more low-frequency unique alleles (trial r in this
case), that latter set of trial conditions can also produce a greater standard
deviation of mean expected heterozygosity across repetitive identical runs
when higher-frequency alleles are involved. That pattern occurs for all
the allele panels, except for trial allele panel B (with the largest number of