Box C

Guidelines for the use of models to estimate pollen-mediated immigration

1. Sample sizes of progeny to detect pollen immigration. To estimate the sample size needed for a given study, one needs to know the exclusion probability of genetic markers, the desired level of gene movement one wishes to detect, and the number of potential donors within the study population. Given known maternity, these estimates can be made for pollen-mediated gene immigration, as estimated by Nason in his pollenGF program (method 3) under three sets of conditions: (a) for pollen-mediated gene movement when maternal plant is known (see Table 1); (b) for seed-mediated gene movement; and (c) for gene movement, regardless of parent which is the combined result of pollen and seed movement. It is useful to note that, in all but very small populations, high exclusion probabilities are essential to keep progeny sample sizes within realistic reach.

2. Number of progeny. For any mother, the number of seeds sampled should exceed total number of potential fathers to use currently available paternity analysis programs for estimating individual male fertilities (e.g. Individual Paternity model). For most species, sampling a few mothers in close proximity will reduce the total number of potential fathers that need to be sampled. This sampling scheme is particularly preferable for multi-site investigations. However, using neighboring mothers may not provide an adequate representation of population dynamics, if there is high variation in patterns of gene flow among females according to their positions within the site. For any model which estimates individual male fertility, many progeny per mother should be collected. But, if you are wish to estimate average male fertility over many females, you might be better off to sample many females and, then, number of seeds would not need to be greater than number of males. Alternatively, if one is interested in estimating gene movement from outside only, the need for seed number to exceed pollen donors is not essential, as long as the outside donor pool represents a large number of fathers.

3. Number of fathers. Estimates of gene movement from outside the area will be unreliable unless fathers represent a random sample of the pollen donor pool consisting of a relatively large number of fathers. Therefore, in order to ensure that an estimate of gene movement is reliable, statistical tests should be conducted to verify that the gene immigration represents a random sample of the global pollen allele frequencies.

4. Correlated matings. Most parental analysis models assume that seeds are sampled at random from the available pollen pool. As an extension of the previous point, it is recommended that one seed per fruit be sampled to avoid non-random sampling of immigration events. For species with pollination mechanisms that promote correlated matings, use of several seeds from one fruit creates high variance in the estimate of gene movement. For the special case of singly-sired fruits, information from correlated matings is desirable and therefore multiple seeds (a full-sib progeny array) from the same fruit should be collected. If you know for sure that correlated matings occur, you can increase precision of paternity assignment.

5. Optimal loci. The best genetic markers are those loci with alleles in equal proportions because those give the highest exclusion probabilities. Even though microsatellite loci hold tremendous promise in terms of allelic diversity, if many alleles are rare and only one is common, they may not be very helpful. For further dicussion of this problem see Smouse and Meagher 1994 and Selvin 1980.

Return to Part I.