Name: Donagh Berry
Title: Quantitative geneticist, Teagasc, the Irish Agriculture and Food Development Authority
Education: PhD, quantitative genetics, University College Dublin and Wageningen University in the Netherlands; MSc, bioinformatics and systems biology, University College Cork, Ireland; BSc, agriculture science, University College Dublin
Teagasc, which means "instruction" in Irish, is the Agriculture and Food Development Authority in Ireland. Overseen and funded by the Ministry of Agriculture, it is the interface between the Irish government, the domestic livestock industry, and individual farmers, and it is the channel through which genomic tools, especially SNP genotyping arrays, have been made available over the past few years.
As a quantitative geneticist within Teagasc, Donagh Berry has many roles, but all of them are focused on implementing genomic selection in Ireland by using SNP arrays to select for better beef and dairy cattle and other livestock to benefit the agriculture sector and, ultimately, the Irish taxpayer. For a relatively small country, Berry says that demand has been high for SNP genotyping services from those who raise cattle. For example, when Teagasc announced that it would be offering genotyping on the Illumina BovineLD BeadChip at the beginning of this year, individual farmers responded by requesting genotyping of 7,000 dairy heifers within the first month of availability.
Berry described his experience in implementing genomic selection in Ireland at a workshop hosted by GeneSeek at the Plant and Animal Genome Conference in San Diego last month. BioArray News spoke with Berry after the conference. Below is an edited transcript of that interview.
Was working in the agbio field something you were aiming to do or something you came in to over time?
Definitely something I came into over time. I am trained as a quantitative geneticist or as a statistical geneticist. Traditionally we [quantitative geneticists] knew absolutely nothing about genomics. We would have been taught very little on that. But because our expertise is in statistics and data handling and databases we were able to adapt. As long as we can understand the terminology, we're fine. Most of us, if not all of us, don't really care to some degree how the data is generated. We don't care about the 454 versus Solexa. All we want is good data. So, I only started getting into genomics since about 2008. But there is a lot of data being generated and we are the ones charged with handling that data.
Do you think it is common in the agriculture research community to have come into genomics later than, say, the human research community?
It is common for quantitative geneticists. We all came in around 2008. Geneticists can be broken down into two disciplines. One is molecular genetics, which are the scientists who generally design and run the laboratory experiments, and then you have the statistical geneticists who don't really care all that much about the biology; they just want to analyze the numbers. We only really took off in about 2007 and 2008. The approach we use is called genomic selection. That was originally proposed back in 2001. The technology though wasn't there at the time to provide us with sufficient molecular genetic information to achieve what we wanted. However, the molecular geneticists have now done that for us, so it's very much a symbiotic relationship.
The [lllumina BovineSNP50] technology was launched in 2008, and then, all of a sudden, we could apply the idea from 2001. We had the genomic tools and they became available at a relatively low cost. There was an off-the-shelf product for someone like me who doesn't know how to do DNA extraction, but could ring up Illumina and say, "I want 10,000 of those arrays. I am going to send you the DNA and you are going to send me back a genotype." And because it is off the shelf, all the data that comes back is of very high quality.
Who told you to adopt these newer tools? Do you have to get permission from your managers to do that or are you free to use what tools you like?
I am free to use whatever tools I like within reason. If I tried to sequence every animal, they'd say, "That's going to cost 10 billion quid, you can't do that." So I have to put a business plan together. It all started in late 2008 when we got a research project to genotype 192 animals. Then I convinced industry to give me sufficient money to do 500 more animals and then I convinced my own organization to give me enough money to do another 500 animals. So at the end of 2008, the start of 2009, I had 945 animals genotyped. Then we implemented genomic selection. For the next 12 months I hassled everyone around the world saying, "Look, I've got these 945 animals. You probably have another 1,000 animals. How about I give you [the data on] my 1,000 and you give me your 1,000, and then the two of us have 2,000."
I did a lot of that with countries like New Zealand, Poland, the UK, Italy and others. At the end of 2009, I had 3,360 genotypes. As of 2011, we have 4,550 animals genotyped. So the whole thing has only cost €500,000 ($662,000), which is very, very cheap, considering that the benefits are €8 million, annually.
What does genomic selection mean for the Irish cattle industry, both dairy and beef cattle? And does it have any ramifications for other livestock research?
There are about 1.05 million dairy cows and 1.1 million beef cows in Ireland. There are about 3.5 million ewes. There is very little pork and poultry, but pigs and poultry around the world are controlled by a small number of large commercial companies. In general, we don't engage in pig and poultry for that reason, because it is full of IP, and we are public. I work in beef, dairy, sheep, and horses.
In dairy, it's all been implemented. The cost of genomic selection is just €29 for the average dairy farmer, using the Bovine LD. That was launched at the start of January, and as of a few weeks ago genotyping had been requested for 7,000 dairy heifers paid by the farmers themselves to do genomic selection. So, pretty good uptake. And that was the first time such an initiative had been launched.
If we keep the same kind of breeding structure, genomic selection can increase genetic gain by at least 50 percent. If we alter the breeding structure, if we use things like embryo transfer and genotype more females and more bulls, you can increase that more dramatically. Generally the international literature says genomic selection can increase genetic gain between 50 and 200 percent.
In the beef [industry], we have started it, but the difficulty with beef is that you do not have large breeds like the Holsteins, so in order to implement genomic selection you have to have [genotypes and phenotypes] for a large population. You have that in the Holstein. In beef, we don't [frequently] use artificial insemination, and therefore we don't have a large number of AI bulls that can be used in a training population. So we have started to genotype natural mating bulls, but even at that it is difficult to get a large number of beef animals so what we are now investigating is across-breed genomic selection. In other words, exploit the DNA information from, say, the Angus in the genomic selection of the Limousin. That is proving a bit difficult at the moment.
In sheep it is a very different situation. First, you have the smaller population size. More importantly, the value of the animal is less. You cannot spend a lot of money on this. For an average farmer, the price of a lamb might be between €70 and €100. So he cannot spend €30 of that to genotype an animal. Therefore, it is a very different equation with the sheep because of the cost of the animal relative to genomic selection.
The technologies are there. The bovine chip is applicable across dairy and beef, but I guess it was really designed for the Holstein, so it probably won't be as good in beef, of course we don't know that. The sheep has a lower-density chip and the simulations suggest that the lower density chip will not allow you to go across breeds, but this is dependent on pedigree structure. So, at the moment, we are not doing any genomic selection research in sheep. But also the breeding infrastructure for sheep in Ireland is not optimum, so we are putting more of our efforts into traditional breeding, getting farmers to record things better, trying to get our genetic evaluations correct and trying to get our breeding schemes correct. Genomics is not useful unless you have the basics correct first.
Do you believe that other industries, like poultry, like pork, as well as crops, will eventually implement genomic selection?
They probably have, a lot of them. They have not traditionally been adopting genomic selection as we know it. Genomic selection is based on SNPs or pieces of DNA across the whole genome. But, for example, in pigs they have been selecting on certain genes … for a long time. The difference with pigs, poultry, as well as corn, is that the population used by the farmer is the first cross between two parental breeds. Why do people do that? The biological rationale is that if you cross two genetically different animals, you get heterosis, a "kick" in performance because the genes of two breeds never really saw each other before and the mixture is better than either of the parental gene copies. But moreover, you as a farmer cannot generate the animals unless you have the two parental lines. So me, as a commercial breeder, I own the parental lines, and I own a continuous business from you, because you are always going to have to buy your replacements from me. Even though the population of pigs in Ireland may be half a million, that's really only coming from a small number of elite animals that is owned by a particular commercial company. So the farmer there does not do any genomic selection, and it's the same in poultry, where the likes of Aviagen are driving the genomic selection of animals. And even though a pig might be worth a hundred euros, their pigs might be worth thousands of euros.
You mentioned before that you already had 7,000 requests for bovine genotyping. How do you reach out to the farmers?
We deal directly with everybody, so last night I had a meeting with farmers. I was talking to around 120 farmers on breeding and on genomic selection. But we tackle it from a number of different areas. We tackle it from the genetic evaluation body, which is the Irish Cattle Breeding Federation. I do the research, I pass the software to those guys, and they implement this. Explaining what I've done to the ICBF can be very technical. Then you go down a level and you talk to breeding companies. That's a little bit less technical but more in terms of how they exploit the new technology to drive their genetic gain. And then you go down to the next level, which may be the [artificial insemination] technicians … to explain how to use this information to select bulls. Then you go down another level to the farmers, and explain how they can use genomic selection with their females. So around 15 percent of my time in Teagasc is talking to stakeholders. Tomorrow night I will be discussing genomic selection with 200 more farmers.
Based on your experience, how would you place Ireland in the context of other countries with large livestock industries that are implementing genomic selection. Are you leading the way?
It depends on how you break it down. Technically, we are not leading the way, but we are not behind, we are in the middle. But we were the second country in the world to implement genomic selection, just a few weeks after the US. That was at the end of February 2009. Exploitation wide, I would say that we are leading the way. The difference with Ireland is that, as far as I know, we are the only country in the world where a farmer can genotype a male animal. The reason why is that I have been genotyping using taxpayers' money. So the Irish people own the genotypes. In the US, they were paid for by commercial companies. Therefore the companies own all the IP. As far as I know, a US farmer cannot genotype a male calf. It has to go through one of the AI organizations. We are very open. You can send us the DNA of an animal, and we will give you a genomic proof. There is a very different system in Ireland.
How have you managed with the data analysis?
Not a problem because I have written all the software. The only problem is finding the time to write better software. I am the only one working on it, and it is about 10 percent of my time. The US has the best brains in the world, and a lot of them [are]working on genomic selection. In Holland or Denmark they might have four to eight people working on it. In Ireland, there's just me. That's why I would say that, technically, we are just in the middle when it comes to genomic selection. It's fine. We are getting by. But we could probably do more sexy things if we had more resources.
A number of arrays have been introduced since the BovineSNP50, including the Bovine HD and Bovine LD. At the GeneSeek workshop you heard that they are developing their own specialized tools for bovine research. What are you going to use?
The specialized one. That has got the [Bovine]LD SNPs on it for imputation and for breeders and well as selection of other causative mutations and markers. For research we already have 4,000 HD chips done. For research you want to have more data, but obviously that comes with a cost. But the farmer doesn't need that amount of information so he can go for the lower cost LD chip.
At the end of the day, it's all about the cost. We have a company in Ireland that does this [called] Wetherbys – they are primarily set up for parentage in horses. But if we put out a tender and GeneSeek comes back with a better price, we'll use GeneSeek.
Is there any way the current SNP chips could be improved?
More content, more information at lower cost. There's more to life than SNPs as well – I would like to see more CNVs on the chips.