2008-09-09

Genomic selection

Last week I have been on a course "Whole genome association and genomic selection" by Ben Hayes. Current course webpage does not (yet) provide course materials (notes and slides), but there are materials from a previous course by Ben - look here.

In a "nutshell" the course was about usage of whole genome data in order to improve (via selection of best individuals) important traits in livestock. Today, we can collect a lot (several thousands), up to 60K for cattle for example) of genotypes with "chip technologies" for a price around 200-300$. After the cleaning of the data, a researcher would estimate effect of each genotype (just additive or maybe even the dominance effect) on the phenotype of individuals. Technically, when we get the genotype data, this data do not represent the exact (functional) genes, but just markers along the whole genome. When we estimate the effect of markers, we try to exploit the possible linkage between the markers and potential genes, which are to a large extent unknown to us. Essential issue with the success of this method is that there is enough linkage disequilibrium between markers and genes. We can achieve that with dense chips. Since there has been an intense selection and strong founder effects in populations of livestock, the re is more linkage disequilibrium and the density of chips in livestock can be lower than in humans. After the estimation of markers effects, we can calculate a genotypic value for any genotyped individual. The essence of this approach has been presented in the paper by Meuwissen et al. (2001). There has been a lot of research/work done since then! There are also quite some important technical details, but this really seems to be a new very hot topic for animal breeding and genetics.

During the course I had an opportunity to disscuss with many people and there seems to be a great intereset for genomic selection in dairy cattle. This is due to the availability of dense chips for cattle and possibility of wide exploitation of semen via artifical insemination. New Zealand, Netherlands, USA are the countries which have already implemented the genomic selection and are already starting to sell the semen of young bulls. The big advantage of genomic selection in dairy cattle is in shortened generation interval. Up to now, potential bulls were preselected from best cows accounting for the breeding values of these cows and inbreeding. Then the semen of this preselected bulls would be used to produce their daughters. Then breeding values of the bulls would be estimated from the phenotypic records (milk yield) of their daughters. Well, the calculation of the breeding values accounts for all possible information from relatives, but the accuracy of breeding values is much dependent on the number of close relatives that have phenotypes. A nice description of these techniques is in the book by Mrode. There are also other books, but the book by Mrode is a very good one for the start. The process from preselecting the bull up to the calculation of breeding values can last around five years.

Besides shortened generation intervals there is also some gain in the accuracy of estimation of genetic values. Up to now animal breeders were mainly exploiting the so called infinitesimall model, where we postulate that there are infinite number of genes and that the effect of gene is negligible. This model is clearly wrong, but has been and still is a very usefull abstraction. Its use has enabled genetic improvement of many economically important traits in agriculture. Since the infinitesimall model is only an abstraction, geneticists were and are still trying hard to find genes that have an effect on various traits. There has been quite a lot of succes in this work, especially for the traits that are under the influence of only few genes. However, for the traits that are continous in the nature there have been only few genes found even though the efforts are substantial. Genomic selection now provides another abstraction. It is again an approximation, but given the abundant genomic data, there are gains in comparison to an infinitesimall model.

With genomic selection animal breeders gain a lot of flexibility and it seems that we will face changes in the organization of breeding programmes. Genomic selection is of course aplicable also for other breeding programmes: beef cattle, sheep, goats, pigs, poultry, ...

No comments: