Improved estimation of aquaculture associated European introgression in a captive breeding program for endangered Atlantic salmon

Abstract The rapid, range-wide decline in Atlantic salmon, Salmo salar , populations is well documented and has led to establishment of captive rearing and breeding programs in order to preserve populations. However, recovery potential may be limited by the inclusion of non-local genotypes, which can be both difficult to detect and quantify. In the genetically unique Inner Bay of Fundy population located in Canada, three Live Gene Bank programs have been established to aid recovery of this endangered conservation unit. Evidence of aquaculture associated non-local (i.e., European) introgression had previously been detected using small panels of microsatellite markers with limited power. Here we show how advances in sequencing and machine learning technologies can support a conservation program. We used machine learning and a corresponding panel of 301 SNPs to estimate individual-level proportions of European ancestry. To assess the degree of introgression in each program and to assess changes over time, fish were randomly selected across several program generations. Estimates were validated by genotyping a subset of individuals on a 220 K SNP array and using established admixture methods. Of the 1741 fish analyzed, only 48 were found to have European ancestry greater than the detection threshold. We found the amount of European ancestry was previously overestimated, and that very few wild-collected founder individuals had large proportions of European ancestry. Moreover, because European ancestry was introduced to Bay of Fundy populations via introgression from aquaculture escapees, these values represent the minimum amount of aquaculture introgression in these captive populations.