Genomic selection (Accordion)

What is genomic selection?

In genomic selection (GS) the genetic potential of an animal (in other words the breeding values, BV) is estimated by taking into account the thousands of markers which are spread across the entire genetic material. In this context a marker is an SNP (Single Nucleotide Polymorphism, pronounced Snip.) SNPs can be determined relatively cost-effectively in the laboratory and even for a young animal – and now even in an embryo.
Once the relationships between the SNPs and the traits that interest us have been determined, breeding values can be calculated with the SNP information from the laboratory (“direct genomic breeding value” or DGV). To improve reliability of these “laboratory breeding values” traditional information sources are then added (breeding value of the parents) – and if available – details on performance and information on progeny and expressed in a combined breeding value as genomic optimized breeding value (GOBV). Genomic breeding values are – with the correct estimate and especially in young animals – more reliable than traditional breeding values (see breeding value reliability). This opens up new breeding related possibilities.

Genetic bases

Genetic material (genome) of living beings is located in the cell nuclei, broken down into individual chromosomes. Bovine cells – apart from the gametes - each contain a total of 2 x 30 chromosomes, half of which originate from the father and half from the mother. The chromosomes contain DNA, a long thread-like molecule that consists of individual bases (four individual bases: A, C, G and T). In the DNA double-strand bases A and C always form a pair as do G and T (base pair, see graphic). The sequence of the bases in the DNA double strand (for example, TCGGATA etc.) form the genetic code.

The entire bovine genome contains around 3 billion base pairs. One section on the DNA strand, which contains the instructions (code) for forming a protein, is called a gene. One gene consists on average of about 4,000 base pairs. In total there are an estimated 30,000 genes in the bovine genome. Genes only comprise about 5% of the total genome. The function of the remaining 95% of the base pairs in the genome is in parts still unknown.

A variation in the genome on one individual base pair (example AAGGTTA and ATGGTTA) is called a Single Nucleotide Polymorphism (SNP, pronounced Snip). SNPs occur very frequently in the genome (in more than 1 million locations), but generally have no effect on a trait. SNPs can be determined relatively easily in the laboratory (see “How are SNPs determined in the laboratory?” and are therefore interesting markers.

What are the benefits of genomic selection?

Genomic selection enables breeding value to be estimated more reliably (cf. “From SNPs to breeding values”) in particular for young animals. Over 70% reliability for calves is possible today. This corresponds to milk in a progeny test with around 25 daughters.

Thanks to a more reliable breeding value

• Breeding animals can be selected with more certainty
• Selection decisions can be made much earlier
• Breeding progress can proceed more quickly

This higher level of reliability and certainty is not restricted to performance traits. All traits benefit from genomic selection and in which the relationship to the SNPs can be developed. Functional traits in particular (longevity, health, etc) are already benefiting considerably from this new technology.

In principle, with reliable genomic breeding values of 70% and more, the question must be asked whether the expensive progeny testing of AI-bulls (costs of over CHF 40,000 per bull) could not be totally eliminated. This would indeed signify a huge cost saving.

However, some type of progeny testing will still be necessary, because

• Genomic selection must still prove itself further. 
• For many breeders 70% reliability is not sufficient 
• Traditional breeding values will continue to be required so that the bases of genomic selection (relationships between the SNPs and the traits of interest) can be maintained

In the meantime the type of progeny testing has been adapted to the new technology in many countries. 

Summary

Genomic selection delivers – and in particular for young animals – more reliable breeding values and therefore a more reliable breeding progress per year along with a considerable cost saving potential.

Some characteristics of an animal – horns / polledness, for example, as well as many hereditary defects – are controlled from one single location in the genome. With other traits, such as yield, size and longevity, many locations in the genome play a role, in addition to environmental influences. Such traits are called quantitative traits. The figure below shows the influence of individual chromosome regions on cow fertility (VanRaden et al, 2008). The higher the deviation the stronger the influence.

What can be clearly identified is that:

• All chromosomes influence fertility
• Some locations of the genome play a much bigger role than otherse

A location in the genome that significantly influences a quantitative trait is called a QTL (quantitative trait locus). The precise location of a QTL on the genome is generally not known. However, markers which are located in proximity to a QTL are more and more often becoming relatively easy to determine. 

However, a marker is only then of benefit if 

• It occurs in different variants (example M and m) and 
• One variant of the marker occurs more frequently with the positive variant (+) of the QTL

The nearer a marker is to a QTL, the bigger the chance that it can assume a representative role for the QTL in the genome analysis. The relationship between markers and QTL can be interrupted from generation to generation and mainly depends on the breed. 

SNPs are good markers as they occur frequently in the genome and their determination can be highly automated in the laboratory.

How are SNPs determined in the laboratory?

SNP-chips are used in laboratory analysis. SNP-chips are available from different manufacturers and in different versions (number of SNPs which are analysed per sample). Standard at the outset for cattle (dairy breeds) was the Illumina® BovineSNP50 BeadChip, with which over 54,000 SNPs were determined simultaneously on 12 samples. There are now others with fewer SNPs called low-density chips (LD-chips) and others with over 777,000 SNP, described as high-density chips (HD-chips). Added to this are chips which, in addition to SNPs, also offer hereditary defect tests (e.g. SMA, Weaver, BLAD, CVM) for estimating breeding values and some tests for certain monogenically inherited characteristics (e.g. red factor, kappa-casein, polledness for certain breeds). Genotypes of LD-chips can be projected through imputing on a higher density (standard chip or even HD-chip). HD-chips are currently only used in research projects because they do not deliver any significant benefit for breeding value estimation.

From SNPs to breeding values

Most SNPs, as determined in the laboratory, tell us nothing initially about the hereditary potential of an animal. Only through statistical methods are the connections between SNPs and the characteristics of interest for breeding exposed. The necessary data often deliver proven bulls whose hereditary patterns are well-known. To this end, many studied proven bulls are required, presumably over a thousand; the more the better. Cows with reliable breeding values can also make a contribution.

Once the relationships have been determined, using SNP studies in combination with the information about their parents, breeding values for young animals can be determined which are significantly more reliable than pure lineage breeding values. How reliable these genomic breeding values are, depends primarily on how well the relationship between SNPs and breeding values can be estimated and which heritability the trait manifests.

How reliable are genomic optimized breeding values (GOBV) in comparison to traditional breeding values (BW)? There have been various studies on this, for example, by Van Doormaal (2009) for a selection of traits on Canadian Holsteins:

Breeding value reliability (coefficient of determination, B%) Holstein Canada (Van Doormaal, 2009)

The inclusion of SNP information, primarily for calves, delivers a significant increase in reliability in comparison to the previous method. But cows can also be assessed more reliably with SNP information than was possible in the traditional way of estimating breeding values based on individual yield and progenitor information. For bulls that have been proven in Switzerland, on the other hand, there is no further increase in reliability to be achieved.

How reliable are genomic optimized breeding values?

Breeding values are an estimate, whether they are genomic optimized, or traditional, breeding values. As an estimate they have a certain reliability. This reliability is generally expressed as a statistical value (coefficient of determination, B) in percent. The reliability takes into account the fact that the breeding value could not be calculated on the basis of an endless number of progeny but could only be estimated on the basis of a restricted amount of information.

Another measurement that describes the reliability of an estimated breeding value is the 95% confidence interval. This expresses the range in which the true breeding value in 95 of 100 cases is around the estimated breeding value.

The number of daughters in pure progeny testing with which the same reliability would be achieved can also indirectly illustrate a breeding value.

The table below – using the breeding value of milk – indicates the confidence interval and a corresponding number of daughters for a stipulated coefficient of determination.

Further explanation

The traditional lineage breeding value of a calf has a reliability of about 35% which corresponds to the reliability of a progeny tested bull with 5 daughters. If the lineage breeding value of a calf is estimated at +800 kg, the true breeding value in 95 of 100 cases is in a range of +800 ±790kg, in other words in a range from +10 kg to +1,590 kg. 

In American studies 70% reliability has already been achieved on a calf for the genomic optimized breeding value of milk. This reliability can only be specified for a proven bull with 24 daughters. If in turn an estimated breeding value of +800 kg is assumed, the true breeding value for this coefficient of determination is in 95 of 100 cases in the range of +263kg +1,337 kg. 

With a traditionally progeny-tested Swissgenetics bull, a reliability of 85 - 90% is achieved through testing. If we assume a reliability of 90%, this corresponds to 91 daughters. With an estimated breeding value of +800kg milk, the true breeding value lies in the range of +490 to +1,110 kg. If the bull is put to secondary use on a large scale the breeding value can be estimated with a reliability up to 99%. 

Conclusion

Genomic optimized breeding values of a calf achieve a reliability of 50% to over 70% depending on the trait and the number of proven bulls used to develop the system. This is significantly higher than a pure lineage assessment but not so high as with a proven AI-bull.

Can genomic breeding values be calculated for all breeds?

Currently in Switzerland genomic breeding values are only calculated for the Braunvieh (inc. Original Braunvieh), Simmental, Swiss Fleckvieh and Holstein (red and black) and these only for herd book animals. There are no genomic breeding values available in Switzerland for other dairy and beef breeds. However, Mutterkuh Schweiz is currently developing genomic selection for the Limousin breed.

When can I have my animal examined?

Examinations are offered by the three big Swiss cattle breeding associations (Braunvieh Schweiz, swissherdbook and Holstein Breeding Association). You can find the details including electronic order placement on their websites www.braunvieh.ch, www.swissherdbook.ch or www.holstein.ch.

How much does an examination cost?

The costs of an examination are comprised of the laboratory cost and the price for calculating the genomic breeding values. Genotyping with the cheaper LD-chip suffices for normal applications. Added to this are the costs for any potential extra tests. Please consult the website of the relevant breeding organisation for current prices. Breeding organisations sometimes also offer price reductions.

What is required for a laboratory examination?

Essentially blood, semen and also hair roots are suitable for determining SNP. Hair roots are now also particularly recommended for twins. In all cases you must follow the notification from your breeding association on how to proceed. A careful sample extraction is hugely important for success. For example, hair root samples are to be taken with the Geneseek hair card.

In what format will I receive the results?

From all breeding associations you will receive a genomic optimized breeding value (GOBV) for (virtually) all traits. These have been official since December 2010 and they replace the traditional breeding values.

How long will it take for me to receive the results?

The samples are shipped according to the instructions of the breeding organisations to Qualitas AG in Zug. From there they are forwarded twice a month – as a rule on the first and third Tuesday of the month – to the laboratory. Normally the genomic breeding values are available 30 - 50 days after sending in the sample. The results are delivered to you in writing by your association.

Are the genomic breeding values stated in official documents?

The genomic optimized breeding values (GOBV) have been official since December 2010 and they replace traditional breeding values. They are identified in Switzerland as follows:

• GA: Direct genomic breeding value (DGV) combined with the traditional lineage breeding value.
• G: DGV combined with the Swiss progeny testing result
• GI: DGV combined with the Interbull breeding value

Do direct genomic breeding values change?

Yes, they can change. Although laboratory values (SNP genotypes) are fixed, the estimated equations with which the genomic breeding values are calculated from the laboratory values are regularly updated.

Will there continue to be progeny testing?

For the moment progeny testing in Switzerland is not for discussion. There will also be a need in the future for a type of verification due to progeny yields.