Limits...
Evaluation of the linkage-disequilibrium method for the estimation of effective population size when generations overlap: an empirical case.

Saura M, Tenesa A, Woolliams JA, Fernández A, Villanueva B - BMC Genomics (2015)

Bottom Line: The N e at the time of the foundation of the herd (26 generations ago) was 20.8 ± 3.7 (average and SD across replicates), while the pedigree-based estimate was 21.This supports the use of the method for estimating N e when pedigree information is unavailable in order to effectively monitor and manage populations and to early detect population declines.To our knowledge this is the first study using replicates of empirical data to evaluate the applicability of the LD method by comparing results with accurate pedigree-based estimates.

View Article: PubMed Central - PubMed

Affiliation: Departamento de Mejora Genética Animal, INIA, Carretera de la Coruña km 7.5, 28040, Madrid, Spain. saura.maria@inia.es.

ABSTRACT

Background: Within the genetic methods for estimating effective population size (N e ), the method based on linkage disequilibrium (LD) has advantages over other methods, although its accuracy when applied to populations with overlapping generations is a matter of controversy. It is also unclear the best way to account for mutation and sample size when this method is implemented. Here we have addressed the applicability of this method using genome-wide information when generations overlap by profiting from having available a complete and accurate pedigree from an experimental population of Iberian pigs. Precise pedigree-based estimates of N e were considered as a baseline against which to compare LD-based estimates.

Methods: We assumed six different statistical models that varied in the adjustments made for mutation and sample size. The approach allowed us to determine the most suitable statistical model of adjustment when the LD method is used for species with overlapping generations. A novel approach used here was to treat different generations as replicates of the same population in order to assess the error of the LD-based N e estimates.

Results: LD-based N e estimates obtained by estimating the mutation parameter from the data and by correcting sample size using the 1/2n term were the closest to pedigree-based estimates. The N e at the time of the foundation of the herd (26 generations ago) was 20.8 ± 3.7 (average and SD across replicates), while the pedigree-based estimate was 21. From that time on, this trend was in good agreement with that followed by pedigree-based N e.

Conclusions: Our results showed that when using genome-wide information, the LD method is accurate and broadly applicable to small populations even when generations overlap. This supports the use of the method for estimating N e when pedigree information is unavailable in order to effectively monitor and manage populations and to early detect population declines. To our knowledge this is the first study using replicates of empirical data to evaluate the applicability of the LD method by comparing results with accurate pedigree-based estimates.

No MeSH data available.


Comparison of estimates of effective population size (Ne) based on linkage disequilibrium with those based on pedigree data, averaged across replicates. Average LD-based Ne estimates across replicates (solid lines; bars represent 1 SD) compared to pedigree-based estimates (dashed lines). Estimates based on LD were obtained estimating (a, b and c) or fixing α to 2 (d, e and f) and ignoring (a and d) or accounting (b, c, e and f) for the sampling effect. The term 1/N is the adjustment for sample size
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4642667&req=5

Fig4: Comparison of estimates of effective population size (Ne) based on linkage disequilibrium with those based on pedigree data, averaged across replicates. Average LD-based Ne estimates across replicates (solid lines; bars represent 1 SD) compared to pedigree-based estimates (dashed lines). Estimates based on LD were obtained estimating (a, b and c) or fixing α to 2 (d, e and f) and ignoring (a and d) or accounting (b, c, e and f) for the sampling effect. The term 1/N is the adjustment for sample size

Mentions: In general, average values of Ne across replicates at the time of the foundation of the herd were lower when α was estimated than when it was fixed. Standard deviations of the Ne estimate across replicates were higher in models where α was fixed (17.5 ± 3.5, 20 ± 3.6, 22.5 ± 3.5 for scenarios a, b, c, respectively, and 26.5 ± 8.5, 26.5 ± 8.5 and 34.5 ± 9.5 for scenarios d, e, f, respectively). All scenarios analyzed provided estimates significantly different from zero for both α (scenarios a, b and c) and Ne. With the exception of model c (Fig. 3c), models where α was estimated from the data led to Ne estimates that slightly increased from the time of the foundation of the herd to approximately six generations ago, when a decline occurred. Models where α was fixed to 2 led to Ne estimates that decreased progressively across generations (Fig. 3d, e, f). The molecular Ne(t) estimates closest to pedigree-based estimates were those obtained under the model where (i) the parameter α was estimated; and (ii) a correction of 1/2n for sample size was imposed (i.e., model b, Figs. 1 and 4b).Fig. 4


Evaluation of the linkage-disequilibrium method for the estimation of effective population size when generations overlap: an empirical case.

Saura M, Tenesa A, Woolliams JA, Fernández A, Villanueva B - BMC Genomics (2015)

Comparison of estimates of effective population size (Ne) based on linkage disequilibrium with those based on pedigree data, averaged across replicates. Average LD-based Ne estimates across replicates (solid lines; bars represent 1 SD) compared to pedigree-based estimates (dashed lines). Estimates based on LD were obtained estimating (a, b and c) or fixing α to 2 (d, e and f) and ignoring (a and d) or accounting (b, c, e and f) for the sampling effect. The term 1/N is the adjustment for sample size
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4642667&req=5

Fig4: Comparison of estimates of effective population size (Ne) based on linkage disequilibrium with those based on pedigree data, averaged across replicates. Average LD-based Ne estimates across replicates (solid lines; bars represent 1 SD) compared to pedigree-based estimates (dashed lines). Estimates based on LD were obtained estimating (a, b and c) or fixing α to 2 (d, e and f) and ignoring (a and d) or accounting (b, c, e and f) for the sampling effect. The term 1/N is the adjustment for sample size
Mentions: In general, average values of Ne across replicates at the time of the foundation of the herd were lower when α was estimated than when it was fixed. Standard deviations of the Ne estimate across replicates were higher in models where α was fixed (17.5 ± 3.5, 20 ± 3.6, 22.5 ± 3.5 for scenarios a, b, c, respectively, and 26.5 ± 8.5, 26.5 ± 8.5 and 34.5 ± 9.5 for scenarios d, e, f, respectively). All scenarios analyzed provided estimates significantly different from zero for both α (scenarios a, b and c) and Ne. With the exception of model c (Fig. 3c), models where α was estimated from the data led to Ne estimates that slightly increased from the time of the foundation of the herd to approximately six generations ago, when a decline occurred. Models where α was fixed to 2 led to Ne estimates that decreased progressively across generations (Fig. 3d, e, f). The molecular Ne(t) estimates closest to pedigree-based estimates were those obtained under the model where (i) the parameter α was estimated; and (ii) a correction of 1/2n for sample size was imposed (i.e., model b, Figs. 1 and 4b).Fig. 4

Bottom Line: The N e at the time of the foundation of the herd (26 generations ago) was 20.8 ± 3.7 (average and SD across replicates), while the pedigree-based estimate was 21.This supports the use of the method for estimating N e when pedigree information is unavailable in order to effectively monitor and manage populations and to early detect population declines.To our knowledge this is the first study using replicates of empirical data to evaluate the applicability of the LD method by comparing results with accurate pedigree-based estimates.

View Article: PubMed Central - PubMed

Affiliation: Departamento de Mejora Genética Animal, INIA, Carretera de la Coruña km 7.5, 28040, Madrid, Spain. saura.maria@inia.es.

ABSTRACT

Background: Within the genetic methods for estimating effective population size (N e ), the method based on linkage disequilibrium (LD) has advantages over other methods, although its accuracy when applied to populations with overlapping generations is a matter of controversy. It is also unclear the best way to account for mutation and sample size when this method is implemented. Here we have addressed the applicability of this method using genome-wide information when generations overlap by profiting from having available a complete and accurate pedigree from an experimental population of Iberian pigs. Precise pedigree-based estimates of N e were considered as a baseline against which to compare LD-based estimates.

Methods: We assumed six different statistical models that varied in the adjustments made for mutation and sample size. The approach allowed us to determine the most suitable statistical model of adjustment when the LD method is used for species with overlapping generations. A novel approach used here was to treat different generations as replicates of the same population in order to assess the error of the LD-based N e estimates.

Results: LD-based N e estimates obtained by estimating the mutation parameter from the data and by correcting sample size using the 1/2n term were the closest to pedigree-based estimates. The N e at the time of the foundation of the herd (26 generations ago) was 20.8 ± 3.7 (average and SD across replicates), while the pedigree-based estimate was 21. From that time on, this trend was in good agreement with that followed by pedigree-based N e.

Conclusions: Our results showed that when using genome-wide information, the LD method is accurate and broadly applicable to small populations even when generations overlap. This supports the use of the method for estimating N e when pedigree information is unavailable in order to effectively monitor and manage populations and to early detect population declines. To our knowledge this is the first study using replicates of empirical data to evaluate the applicability of the LD method by comparing results with accurate pedigree-based estimates.

No MeSH data available.