The Journal of Heredity 2001:92(3)
© 2001 The American Genetic Association 92:243-247
The Aunt and Uncle Effect: An Empirical Evaluation of the Confounding Influence of Full Sibs of Parents on Pedigree Reconstruction
From the Gene Conservation Laboratory, Alaska Department of Fish and Game, 333 Raspberry Rd., Anchorage, AK, 99518-1599 (Olsen), Marine Molecular Biotechnology Laboratory, University of Washington, Seattle, Washington (Olsen, Britt, and Bentzen), and Washington Department of Fish and Wildlife, Olympia, Washington (Busack). J. Britt is currently at the Department of Biology, University of Oregon, Eugene, Oregon.
Address correspondence to Jeffrey B. Olsen at the address above or e-mail: jeff_olsen{at}fishgame.state.ak.us.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionSummaryReferences |
|---|
This study used simulations and a known two-generation pedigree of chinook salmon (Oncorhynchus tshawytscha) to evaluate the effect of full sibs of parents on pedigree reconstruction. Parentage analysis was conducted on 100 parent pair-offspring relationships from pedigrees with unrelated (simulation) and related (chinook salmon) candidate parents. Parentage assignment success for the chinook salmon was lower than in the simulated populations. For example, the six most variable loci (mean HE = 0.87) provided a mean of 97% unambiguous assignments in the simulated population and 67% unambiguous assignments for the chinook salmon. Estimates of the pairwise relatedness coefficient (
xy) for most nonexcluded false parents and true parents of chinook salmon offspring exceeded 0.50. These results support the conclusion that closely related candidate parents decrease the power of genetic markers for pedigree reconstruction based on exclusion. Ambiguous parentage may be resolved using single parent- and parent pair-offspring likelihood analysis, however, these methods should be used with caution and they are not replacements for using more loci when many candidate parents are full sibs. |
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryReferences |
|---|
There is growing interest among biologists and ecologists in the use of genetic markers to resolve genealogic relationships in wild and captive populations. Some applications include studying mating behavior and reproductive success (Foote et al. 1997; Primmer et al. 1995), estimating heritability of economically and ecologically important traits (Mousseau et al. 1998), and comparing progeny performance and rearing strategies in captive populations (Estoup et al. 1998; O'Reilly et al. 1998).
When estimating the resolving power of genetic markers for pedigree reconstruction it is generally assumed that candidate parents are unrelated (Chakraborty et al. 1988). If this assumption is invalid, then the resolving power is overestimated. For example, Double et al. (1997) show that paternity assignment success based on exclusion alone declines when candidate males are related. Few studies, however, have examined how closely related candidate parents of both genders affect pedigree reconstruction when knowledge of both parents is desired. This potential error is particularly relevant when estimating reproductive success of captive brood stock composed of groups of closely related individuals and in natural populations with large variance in family size.
The two methods typically used to assign parentage with genetic data are exclusion and likelihood analysis. If only one candidate parent pair is genetically compatible with an offspring, then the assignment is unambiguous and based on exclusion alone. If multiple pairs are genetically compatible, then likelihood analysis is used to infer parentage. The theoretical basis for parent pair-offspring likelihood analysis is well described (e.g., Meagher and Thompson 1986, 1987), however, simulation and empirical studies have focused primarily on refining and testing single parent-offspring likelihood analysis. For example, Marshall et al. (1998) and Slate et al. (2000) use simulation and known pedigrees to demonstrate the accuracy of likelihood-based single parent inference. It is not clear, however, if these results apply to likelihood-based parent pair inference. For example, if knowledge of both parents is required and their identity is initially unknown then both likelihood methods (single parent-offspring or parent pair-offspring) may be used to infer relationships and the resulting pedigrees may differ (e.g., Prodöhl et al. 1998). An empirical evaluation of these two likelihood methods could help identify which one most accurately resolves pedigrees, particularly when some candidate parents are related.
We used computer simulations and empirical data to examine how full sibs of both parents and of both genders affect pedigree reconstruction. Our goals were twofold: (1) to quantify the decrease in parentage assignment success resulting from the presence of closely related candidate parents in a known chinook salmon pedigree; (2) to compare single parent-offspring and parent pair-offspring likelihood methods for resolving pedigrees when nonexcluded candidate parents are full sibs.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryReferences |
|---|
Chinook Salmon Pedigree, DNA Preparation, and Genotyping
Parentage analysis was performed on a known chinook salmon pedigree from a captive brood stock program in Washington State (Smith and Wampler 1995). One hundred parent pair-offspring relationships from 18 families were selected to achieve a representative range in family sizes (3–11 offspring per family). Between 2 and 11 full-sib relatives of each true parent were among the candidate parents (48 males and 54 females).
Tissue samples from adults (caudal fin) and progeny (whole fish) were preserved in 100% ethanol. Total genomic DNA was isolated from 20–30 mg of tissue using procedures based on those for the Gentra Systems (Minneapolis, MN) Puregene DNA isolation kit.
Fourteen microsatellite loci were scored for each individual (Table 1). Polymerase chain reaction (PCR) was carried out in 10 µl volumes [10 mM Tris-HCl pH 8.3, 50 mM KCl, 1.5 mM MgCl2, 0.2 mM each dNTP, 0.5 units Taq DNA polymerase (Promega, Madison, WI), 0.05–0.35 µM each primer, and 100 ng DNA template] using a Perkin-Elmer model 9600 thermocycler. Four multiplexed groups of loci were chosen using methods described by Olsen et al. (1996).
|
Microsatellites were size fractionated using an Applied Biosystems Inc. (ABI) 373A automated DNA sequencer operated in GeneScan mode (ABI 1993). Allele scoring and tabulation of data for importing into statistical software was performed with Genotyper software, version 2.0 (ABI 1996).
Simulated Pedigree
A computer simulation program was used to create 100 parent pair-offspring relationships from a population of unrelated candidate parents. Forty-eight male and 54 female genotypes were created from a random sample of a gamete pool generated from the chinook salmon allele frequency data. One hundred progeny genotypes were created by drawing a male and female parent at random and selecting one of two alleles at random from each locus from each parent. This process was repeated 1,000 times using 4, 6, 8, 10, 12, and 14 loci.
Exclusion Probability and Relatedness
The average exclusion probability for a single unrelated parent-offspring pair was estimated for each locus (PE; Appendix 2 of Marshall et al. 1998) and for all loci (PE(C); equation 3 of Chakraborty et al. 1988) assuming both parents are unknown. The relatedness statistic (rxy) was estimated for each pair of adult chinook salmon and for each group of full sibs using the relatedness equation of Queller and Goodnight (1989) implemented in the computer program RELATEDNESS version 5.0.1 (Goodnight and Queller 1997). Estimates of relatedness (xy) were made using 14 loci, regardless of the number of loci used for parentage analysis. Allele frequency data for the exclusion probability and relatedness equations was estimated from the source population of the chinook salmon captive brood stock.
Parentage Analysis
Parentage analysis was conducted on both pedigrees using 4, 6, 8, 10, 12, and 14 loci included in descending order of PE. All possible crosses (2,592) were considered as candidate parent pairs. Offspring of the simulated pedigree were assigned parentage using exclusion alone because the percent unambiguous assignments exceeded 97% when six or more loci were used. Offspring of the chinook salmon pedigree were assigned parentage using exclusion, single parent-, and parent pair-offspring likelihood analysis. The computer program PROBMAX (Danzmann 1997) was used to identify nonexcluded parent pairs. For offspring with multiple nonexcluded parent pairs a parent pair-offspring (PPO) log-likelihood ratio (LOD) was computed for each triplet using equation 4 of Meagher and Thompson (1986). Parentage was assigned to the parent pair providing the highest PPO LOD score and the assignment was counted as correct if the true parent pair had the highest score. A single parent-offspring (SPO) LOD was also computed for nonexcluded candidate parents using the computer program CERVUS (Marshall et al. 1998), assuming a genotyping error rate of zero. The male and female with the highest SPO LOD scores were identified as the most likely parents and the offspring was counted as correctly assigned if the true parents had the highest scores.
Assignment success was defined generally as the percentage of offspring assigned to their true parent pair. Assignment success was computed first using exclusion (unambiguous parentage) and then using PPO likelihood analysis (inferred parentage) on the remaining unassigned offspring. For the computer simulation assignment success was the mean from 1,000 pedigrees.
The methods above considered all possible crosses in order to mimic the common situation in studies of natural populations where knowledge of mating pairs is not available. However, in captive breeding programs such as this one, the mating scheme is often known. To evaluate parentage assignment success in the situation where the mating scheme is known, PROBMAX (Danzmann 1998) was used to assign parentage in the captive chinook salmon stock given the 134 matings known to have occurred. The results of this approach were compared to the results above that considered 2,592 possible breeding pairs.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryReferences |
|---|
Exclusion Probability and Relatedness
The average exclusion probability (PE) for each locus for a single parent-offspring pair ranged from 0.152 (Ots1) to 0.768 (Ots100) (Table 1). The average exclusion probability for all loci, PE(C), exceeded 0.999. Estimates of average relatedness (
) among chinook salmon parents and their full-sib relatives ranged from 0.267 to 0.767 (average 0.467). All estimates of relatedness were significantly greater than zero based on the 95% confidence interval generated from a jackknife sample of all loci.
Parentage Analysis
The parentage assignment success was always lower for the chinook salmon pedigree than for the simulated pedigrees (Figure 1A). For example, the six most informative loci (PE(C) = 0.995) provided a mean of 97% (1,000 iterations) unambiguous assignments for the simulations and 67% unambiguous assignments for the chinook salmon. The percentage of chinook salmon offspring with unambiguous parentage increased as loci were added, but did not exceed 92% at 14 loci. Both SPO and PPO likelihood analysis increased assignment success for the chinook salmon. Nevertheless, when 12 or fewer loci were used these improvements were not sufficient to meet expectations based on the simulated populations (Figure 1A). Finally, the increase in assignment success varied depending upon the method of likelihood analysis used. Neither method consistently outperformed the other.
|
) and chinook salmon pedigree (
, 
, 
). Parentage analysis was conducted using exclusion for the simulated pedigrees (The mean number of nonexcluded parent pairs (MPPs) was always greater for the chinook salmon than for the simulations (Figure 1B). The mean estimate of pairwise relatedness (
) for nonexcluded false parents and true parents of chinook offspring exceeded 0.5 (the expectation for full sibs) at six loci and increased further as additional loci were used for parentage analysis (Figure 1B). By contrast, when only four loci were used there was a noticeable drop in mean relatedness ( = 0.35), an increase in the standard deviation of r, and a steep increase in MPP (Figure 1B). There was also a large increase in the difference between the chinook salmon MPP and the simulation population MPP. These differences reflect the fact that when only four loci were used, the nonexcluded false parents included individuals related and unrelated (xy = 0.00) to the true parents. The parentage assignment success varied between chinook salmon families (Table 2). Family AA1 always had more possible parent pairs than other families, including those families with a similar number of sampled progeny (AB3, W2). The mean of relatedness estimates for true parent/false parent pairs in family AA1 were always greater than 0.50 and was generally higher than in other families.
|
Finally, knowledge of the breeding pairs greatly improved assignment success in the chinook salmon population by reducing the number of possible parent pairs to 134. Assignment success for 100 progeny was 95% (4 loci), 97% (6 loci), 99% (8 loci), and 100% (10 or more loci). All assignments were unambiguous and likelihood analysis failed to identify the true parent pair in the few instances where multiple parent pairs were not excluded.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryReferences |
|---|
Reduced Assignment Success Due to Closely Related Candidate Parents
The parentage assignment results for the simulated and chinook salmon pedigrees show that the presence of full-sib candidate parents can substantially decrease the exclusionary power of genetic markers. These results contrast with the simulation results of Marshall et al. (1998), who show that the presence of related candidate males has a minor effect on paternity assignment if likelihood analysis is used to infer parentage. The difference in the two studies is likely due to the following reasons. First, Marshall et al. (1998) conducted a paternity analysis and only male relatives of the true father were among the candidate fathers. In the present study, the identities of both parents are unknown so full-sib relatives of both parents, and of both genders, are among the candidate parents. Second, Marshall et al. (1998) considered relatedness estimates only as high as 0.50 for full-sib males. Figure 1B suggests higher estimates of relatedness should be considered to account for full- sib pairs that share more alleles than average. For example, the highest
xy value for a pair of full-sib adult chinook salmon is 0.809. Values of xy greater than 0.50 are possible under two scenarios: stochastic variation in observed allele sharing because relatively few loci are sampled, and actual high relatedness because of common ancestry. Because the average relatedness among all full-sib adult chinook salmon in this study is very close to 0.50 ( = 0.467) we believe a high relatedness estimate (xy) for a pair of full sibs is a result of stochastic variation. Third, Marshall et al. (1998) considered only five close relatives. This total may be to low for some studies; a larger number of relatives increases the probability of very similar full-sib genotypes. In the present study, true parents are related to as many as 11 other candidate parents (Table 2).
Nonzero variance in reproductive success implies that candidate parental populations will be made up of families with varying numbers of siblings. The results presented here show that parentage assignment success for progeny whose parents are part of large families are often, but not invariably, lower than for progeny derived from smaller families. This is evident for families AA1 and G2 (Table 2). Both families have a parent(s) with 9 or 10 relatives in the candidate parent population, and both families have more than one nonexcluded parent pair at 14 loci. In contrast, progeny whose parents have fewer relatives in the candidate parent population can have low assignment success if these sibling candidate parents have high relatedness. This is evident for family AB3 at 14 loci (Table 2). The single false parent, a male, is one of only five full- sib relatives of the true male parent and the estimate of relatedness for the pair is high (xy = 0.66). While these results represent only a single population sample, the data emphasize the loss of resolution that can occur due to variance in reproductive success and high parental relatedness.
Parentage Assignment Using Likelihood Analysis
This study provides empirical support for the use of likelihood analysis to infer parentage of progeny with multiple compatible parent pairs. Nevertheless, Figure 1A reveals two important considerations regarding the use of this approach to resolve pedigrees. First, if the population has many related candidate parents then the use of likelihood analysis does not replace the need for more loci to achieve a level of assignment success comparable to that for a population with unrelated candidate parents. In this study 10 loci are required to exceed 95% assignment success for the chinook salmon pedigree when both exclusion and PPO likelihood analysis are used. Just six loci are required to exceed 95% assignment success, using exclusion alone, for the simulated populations with unrelated candidate parents. Second, when the number of progeny with unambiguous parentage is small then likelihood analysis may result in an unacceptable level of incorrect assignments. For example, using six loci just 67 of the 100 chinook salmon progeny in this study are assigned unambiguous parentage. Of the remaining 33 progeny, 17 and 24 are incorrectly assigned parentage using SPO and PPO likelihood analysis, respectively. In contrast, 85 of the chinook salmon progeny are assigned unambiguous parentage using 10 loci. Of the remaining 15 progeny, 8 and 3 are incorrectly assigned parentage using SPO and PPO likelihood analysis, respectively. These results suggest that likelihood analysis should be used with caution and it is not a replacement for using more loci to resolve pedigrees with many full-sib candidate parents.
The results of the empirical comparison of PPO and SPO likelihood analysis are equivocal. Figure 1A indicates that, for many correctly assigned progeny, the most likely single parents of each gender form the most likely parent pair as suggested by Meagher and Thompson (1986). Nevertheless, this is not the case for all progeny because, regardless of the number of loci used, the two methods always differ in the number of correct assignments. Similarly, Prodöhl et al. (1998) showed that just 51% of candidate mothers and 30% of candidate fathers had both the highest SPO and PPO LOD scores in a population of armadillos. Prodöhl et al. (1998) could not assess success of the two approaches, however, because they had no a priori knowledge of true relationships. In this study we quantify the success of each method using the known chinook salmon pedigree and show that neither method provides consistently higher assignment success when the number of loci vary. This result suggests that, as a conservative measure, both likelihood methods should be used and parentage should be assigned to only those progeny for which SPO and PPO likelihood provide the same parent pair. To minimize the number of incorrect assignments using this approach, more loci should first be used to decrease the number of progeny with ambiguous parentage.
Genotyping Precision and Mutations
Two other factors that can potentially influence the success of pedigree reconstruction are genotyping error and mutation. Results obtained by O'Reilly et al. (1998) suggest that microsatellite genotyping error is about 2–3% per allele scored, but the affect of mutation is negligible. We found no evidence of genotyping error or mutation when reconstructing the known chinook salmon pedigree. We believe the methods used to score genotypes in this study, while not eliminating the risk, greatly reduces risk of some sources of genotyping error. For example, the use of multiplex PCR reduces sample handling; scoring the same individual as a control on each gel helps identify sizing errors due to poor quality gels; the use of in-lane sizing standards improves allele sizing precision and allows for noncontiguous allele size categories that help reveal new alleles or missized alleles. Nevertheless, the probability of genotyping error increases with the size of the pedigree and this should be considered when conducting large-scale parentage analysis using exclusion or likelihood analysis.
|
Summary |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryReferences |
|---|
The presence of closely related candidate parents decreases the resolving power of genetic markers for pedigree reconstruction. The size of decrease, however, is dependent on the number of full-sib relatives of each true parent, the degree of relatedness, and the desired parentage information (single parent or parent pair). Ambiguous parent pair-offspring relationships resulting from closely related candidate parents can be remedied by first adding loci and then using single parent- and parent pair-offspring likelihood analysis.
|
Acknowledgments |
|---|
We thank Chris Marlowe, Sewall Young, Anne Marshall, and other staff of the Washington Department of Fish and Wildlife. This project was funded by Saltonstall-Kennedy grant number NA76FDO299.
|
Footnotes |
|---|
Corresponding Editor: Robert Angus
Received September 7, 2000
Accepted January 15, 2001
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionSummaryReferences |
|---|
-
ABI (Applied Biosystems Inc.), 1993. GeneScan 672 software users manual, revision A. Foster City, CA: Perkin- Elmer.
ABI (Applied Biosystems Inc.), 1996. Genotyper 2.0 users manual. Foster City, CA: Perkin-Elmer.
Banks MA, Blouin MS, Baldwin BA, Rashbrook VK, Fitzgerald HA, Blankenship SM, and Hedgecock D, 1999. Isolation and inheritance of novel microsatellites in chinook salmon (Oncorhynchus tshawytscha). J Hered 90:281–288.
Chakraborty R, Meagher TR, and Smouse PE, 1988. Parentage analysis with genetic markers in natural populations. 1. The expected proportion of offspring with unambiguous paternity. Genetics 118:527–536.
Condrey MJ and Bentzen P, 1998. Characterization of coastal cutthroat trout (Oncorhynchus clarki clarki) microsatellites, and their conservation in other salmonids. Mol Ecol 7:787–788.[Medline]
Danzmann RG, 1997. PROBMAX: a computer program for assigning unknown parentage in pedigree analysis from known genotypic pools of parents and progeny. J Hered 88:333.
Double MC, Cockburn A, Barry SC, and Smouse PE, 1997. Exclusion probabilities for single-locus paternity analysis when related males compete for matings. Mol Ecol 6:1155–1166.
Estoup A, Gharbi K, SanCristobal M, Chavelet C, Haffray P, and Guyomard R, 1998. Parentage assignment using microsatellites in turbot (Scopthalamus maximus) and rainbow trout (Oncorhynchus mykiss) hatchery populations. Can J Fish Aquat Sci 55:715–723.
Foote CJ, Brown GS, and Wood CC, 1997. Spawning success of males using alternative mating tactics in sockeye salmon, Oncorhynchus nerka. Can J Fish Aquat Sci 54:1785–1795.
Goodnight KF and Queller DC, 1997. Relatedness version 5.0. Distributed by author. Houston, TX: Keck Center for Computation Biology, Rice University.
Marshall TC, Slate J, Kruuk LEB, and Pemberton JM, 1998. Statistical confidence for likelihood-based paternity inference in natural populations. Mol Ecol 7:639– 655.[Medline]
Meagher TR and Thompson E, 1986. The relationship between single parent and parent pair genetic likelihoods in genealogy reconstruction. Theor Popul Biol 29:87–106.
Meagher TR and Thompson E, 1987. Analysis of parentage for naturally established seedlings of Chamaelirium luteum (Liliaceae). Ecology 68:803–812.
Mousseau TA, Ritland K, and Heath DD, 1998. A novel method for estimating heritability using molecular markers. Heredity 80:218–224.
Nelson RJ and Beacham TD, 1999. Isolation and cross species amplification of microsatellite loci useful for study of Pacific salmon. Anim Genet 30:228–229.[ISI][Medline]
Nelson RJ, Beacham TD, and Small MP, 1998. Microsatellite analysis of the population structure of a Vancouver Island sockeye salmon (Onchorhynchus nerka) stock complex using nondenaturing gel electrophoresis. Mol Mar Biol Biotechnol 7:312–319.[ISI][Medline]
Olsen JB, Bentzen P, and Seeb JE, 1998. Characterization of seven microsatellite loci derived from pink salmon. Mol Ecol 7:1087–1089.[Medline]
Olsen JB, Wenburg JK, and Bentzen P, 1996. Semiautomated multilocus genotypic of Pacific salmon (Oncorhynchus spp.) using microsatellites. Mol Mar Biol Biotechnol 5:259–272.[ISI][Medline]
O'Reilly PT, Herbinger C, and Wright JM, 1998. Analysis of parentage determination in Atlantic salmon (Salmo salar) using microsatellites. Anim Genet 29:363–370.
Primmer CR, Moller AP, and Ellegren H, 1995. Resolving genetic relationships with microsatellite markers: a parentage testing system for the swallow Hirundo rustica. Mol Ecol 4:493–498.[Medline]
Prodohl PA, Loughry WJ, McDonough CM, Nelson WS, Thompson EA, and Avise JC, 1998. Genetic maternity and paternity in a local population of armadillos assessed by microsatellite DNA markers and field data. Am Nat 151:7–19.
Queller DC and Goodnight KF, 1989. Estimating relatedness using genetic markers. Evolution 43:258–275.
Scribner KT, Gust JR, and Fields RL, 1996. Isolation and characterization of novel salmon microsatellite loci: cross-species amplification and population genetic applications. Can J Fish Aquat Sci 53:833–841.
Slate J, Marshall T, and Pemberton J, 2000. A retrospective assessment of the accuracy of the paternity inference program CERVUS. Mol Ecol 9:801–808.[Medline]
Small MP, Beacham TD, Withler RE, and Nelson RJ, 1998. Discriminating coho salmon (Oncorhynchus kisutch) populations within the Fraser River, British Columbia using microsatellite DNA markers. Mol Ecol 7:141–155.
Smith CJ and Wampler P, 1995. Dungeness river chinook salmon rebuilding project progress report 1992–1993. Northwest fisheries resource bulletin. Project report series no. 3. Olympia, WA: Department of Fish and Wildlife.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  E. C. Anderson and J. C. Garza The Power of Single-Nucleotide Polymorphisms for Large-Scale Parentage Inference Genetics, April 1, 2006; 172(4): 2567 - 2582. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||