Management of Genetic Diversity in Conservation Programs using Genomic Coancestry

Resumo

A main objective in conservation programs is to maintain genetic diversity, and the most efficient management strategy to achieve it is to apply the Optimal Contributions method. This method optimizes the contributions of breeding candidates by minimizing the global weighted coancestry. This leads to the highest levels of genetic diversity, when measured as expected heterozygosity, and to an effective control of the increase of inbreeding. The fundamental parameter of the method is the coancestry matrix which, traditionally, has been obtained from pedigree data. The current availability of genome-wide information allows us to estimate coancestries with higher precision. However, many different genomic coancestry measures have been proposed and it is unknown which measure is more efficient to minimize the loss of genetic diversity. Thus, the general aim of this thesis was to investigate the efficiency of different genomic coancestry matrices in the management of conserved populations when the Optimal Contributions method is applied to maximize genetic diversity. The matrices compared were those based on: i) the proportion of shared alleles (CSIM); ii) deviations of the observed number of alleles shared by two individuals from the expected number (CL&H); iii) the realized relationship matrix obtained by VanRaden's method 1 (CVR1); iv) the realized relationship matrix obtained by VanRaden's method 2 (CVR2); v) the realized relationship matrix obtained by Yangs method (CYAN); and vi) identical by descent segments (CSEG). Results for a single generation using thousands of SNP genotyped in individuals from a farm turbot population, showed large differences in the magnitude of the six coancestry coefficients. Moreover, pairwise correlations were those between coefficients greatly varied (especially for self-coancestry). The lowest correlations between CSIM, CL&H or CSEG and CVR2 or CYAN. Management with matrices based on the proportion of shared alleles or on segments (CSIM, CL&H and CSEG) retained higher variability than those based on realized genomic relationship matrices (CVR1, CVR2 and CYAN). As expected, maximizing heterozygosity pushed alleles toward intermediate frequencies. However, moving allele frequencies away from initial frequencies may be undesirable as particular adaptations to the environment can be lost. Stochastic simulations were used to investigate the efficiency of CL&H and CVR2 in the management of an undivided population across 50 generations and both matrices were compared not only in terms of the genetic diversity maintained but also in terms of the associated changes in allele frequencies across generations. The use of CL&H in the Optimal Contribution method resulted in a higher genetic diversity but also in a higher change of allele frequencies than the use of CVR2. The differences between strategies were reduced when only SNPs with a minimum allele frequency (MAF) above a particular threshold (MAF > 0.05 and MAF > 0.25) were used to compute CL&H and CVR2 as well as when the Optimal Contributions method was applied in populations of smaller sizes (N = 20 vs N = 100). The evaluation of CL&H and CVR2 was extended to subdivided populations, also via computer simulations. When populations are subdivided into different breeding groups, it is possible to give different weights to the within- and between-subpopulation components of genetic diversity. When a higher weight is given to the within-subpopulation component, the levels of inbreeding can be restricted. In this scenario, the use of CL&H was the best option for managing subdivided populations as it maintained more global diversity, led to less inbreeding and to changes in frequencies similar to those observed when using CVR2 when a large weight was given to the within-subpopulation term.