Unlike most recent developments in the application of advanced molecular genetics in animal breeding, Genomic Selection promises to be accessible to relatively modest breeding programmes. This is due in large part to it being less reliant on extensive infrastructure such as separate family tanks and whole-lifetime trait recording.
Whilst only a handful of aquaculture breeding programmes worldwide have deployed the technology so far, it is already clear that uptake is likely to accelerate in the near future. In large part this is because there are features of Genomic Selection that will be of particular value in aquaculture breeding, and conversely there are aspects of aquaculture breeding that will be particularly amenable to Genomic Selection.
Traditional animal breeding relies on using observable physical characteristics, or phenotypes, to identify the best individuals from which to breed. Where an animal’s pedigree is known and phenotypic information is available on its siblings or parents, this data can be combined to assign an Estimated Breeding Value (EBV) to each animal.
This process is necessarily very slow as many traits, such as harvest weight, cannot be measured until a fish or shellfish reaches maturity, or even after harvest in the case of, for example, fillet yield.
Moreover, many traits, most obviously disease resistance, cannot be measured on candidate broodstock at all, with predictions reliant on the performance of siblings. For this reason most advanced aquaculture breeding programmes are family-based and hence able to make selection decisions only on between family differences. Any within family variation is lost to the programme. Since siblings share only 50% of their genes, selection decisions tend to be no more than 50% accurate. Sib based prediction is therefore, necessarily very inefficient and genetic progress falls far short of what is actually possible.
Marker Assisted Selection
Marker Assisted Selection (MAS) is a technique that goes some way to addressing these difficulties. The technology has been widely used in animal and crop breeding, and although Genomic Selection differs fundamentally from MAS, we can better understand how Genomic Selection works by first considering Marker Assisted Selection.
Molecular genetic markers are polymorphisms or DNA sequence variants that can be used as landmarks in the genome. Locating a gene (or genes) involved in traits of interest for selection is an enormously complex task, but fortunately we can predict which individuals carry a favourable gene variant without any knowledge of the function of that gene or where it is in the genome by using genetic markers. This is possible because of a phenomenon known as linkage: quite simply, marker variants that are physically close (linked) to a gene of interest tend to be co-inherited (they are associated), whilst those that are distant will be inherited independently. Statistical tests are used to detect these associations between molecular genetic marker variants and traits of interest.
A MAS example
Suppose we were to genotype the mortalities and survivors of a disease outbreak: we would expect to find that the alternative variants of molecular genetic markers were equally common in both groups. But if we found one variant was overwhelmingly more common in the survivors and very rare amongst mortalities, then put the Champagne on ice, we have identified a marker variant that is genetically linked to our trait! We can now genotype our candidate broodstock and identify the ones that carry the favourable variant and we have doubled our rate of genetic progress.
Marker Assisted Selection has been widely used in animal breeding and though there are few examples in aquaculture, the use of MAS to select for resistance to IPN in Atlantic salmon breeding, has been one of the most successful examples of MAS in animal breeding on record.
In plants, many traits – in particular resistance to fungal diseases – are determined by a single gene and MAS has been more widely used in crop breeding for this reason.
Herein lies the limitation of MAS: it can only help us if our trait is controlled by one, or a very small number of genes with a very large effect (known as quantitative trait loci [QTL]). In practice most complex traits in animals are controlled by many, perhaps hundreds of genes, variation in each contributing just a few percept of the total variation we see. Even where traits of interest are controlled by single genes, we cannot realistically use MAS for more than a handful of traits simultaneously.
Advantages of Genomic Selection
Genomic Selection (GS) develops the idea of Marker Assisted Selection in a totally different direction. Instead of focussing on a small number of significant genes, GS uses the most abundant kind of molecular genetic marker, called SNP’s (single nucleotide polymorphisms) – variants present as single DNA base differences between individuals. SNP analysis exploits a rapidly developing technology – the use of SNP chips, to genotype an individual at thousands, even hundreds of thousands of SNPs simultaneously. At such marker densities, almost every gene will be in at least partial linkage to one or more SNPs.
In Genomic Selection, both phenotypes and genotypes are measured in one population, known as the ‘training’ or ‘discovery’ population and the association between phenotype and genotype is calculated. Comparison with the candidate population genotypes allows calculation of a Genomic Estimated Breeding Value (GEBV) based on its genotype alone. The GEBV is essentially the sum of the effects of all the SNPs, and the power of GS lies in the ability to include the large number of very small associations and thus potentially capture 100% of the genetic variation in a population, not just the effects of the most significant genes.
Moreover GEBV’s are assigned at the level of individual, not family, so we can now identify the highest merit individuals within families.
Another appealing feature is that GEBV’s are self-validating: we can easily divide our training population in two and ‘pretend’ we do not know the phenotype of one half, then compare the predicted GEBV with the true value, to estimate the accuracy of our predictions.
The huge opportunities presented by GS come with considerable further challenges: the large numbers of markers used in GS are not available for all species – although with costs reducing all the time, isolation of markers will soon be within reach for most.
In particular the technique is so new that the algorithms for calculating GEBVs are still being refined and there are significant computational challenges involved in analysis of such vast datasets. Nevertheless, the potential gains are so appealing we should expect to see this exciting technology being applied across aquaculture breeding programmes at an accelerating pace.
The future is no longer what it used to be!
— Halina Sobolewska
Dr Halina Sobolewska is Technical Director of Noahgene Ltd, a genetics services and consultancy company in the UK. You can reach her at: Halina@noahgene.com