Dr Ivan Pocrnic on applying mega-scale data sets to animal breeding
Developing a niche research area in statistical genetics to tackle the challenge of enabling sustainable livestock in a changing world.
Dr Ivan Pocrnic is a Career Track Fellow who uses statistical genetics to interpret large data sets and apply the outcomes in livestock breeding. Dr Pocrnic explained how he applies his skills to support breeders and farmers in managing their livestock, squeezing his data until it tells him what he wants to know, and the challenges and opportunities in his field.
How did you become interested in animal bioscience?
I studied a BSc in agricultural economics at the University of Zagreb in Croatia. During my final year I took an elective course in animal genetics and breeding, and the professor asked if I would like to switch careers. That led me to an MSc in Animal Genetics and Breeding, also at Zagreb.
I became drawn to this because I could see how statistics and data science could be applied to biological problems - I found a niche in statistical genetics. But I had no background in physiology, reproduction or other fundamental biology, so I had to study in my own time to catch up with other people on the masters.
What brought you to Roslin?
After my masters, I got an internship in the Croatian Agricultural Agency, and afterwards was recommended for a PhD with the distinguished scientist Professor Ignacy Misztal – he developed efficient computer programmes applying statistical models to livestock, which are used all over the world. He invited me for a summer course at the University of Georgia, US – a state-of-the-art place to do statistical genetics applied to livestock. I stayed for my PhD, then a postdoc, and then was offered a research scientist position.
At this time, Professor Gregor Gorjanc was starting his group at the Roslin Institute and looking for a core scientist, and I was looking for a new challenge, and I moved to Roslin.
Roslin offers the best of both worlds. My wife and I had visited Edinburgh years before and fallen in love with it.
I enjoy a good balance of a city I really like, plus a well-known place to do quantitative genetics research. I was lucky to work with Professor Gorjanc and was able to do my own projects as well.
What are your main research interests?
I develop novel breeding methodologies and tools using large data sets – at the interface of theoretical quantitative genetics and practical animal breeding.
My main expertise is applying genomic selection on mega-scale datasets for various species such as pig, chicken, or cattle.
What sort of data do you work with?
We use genomic data, such as DNA markers, together with various information from breeding organisations, for example, pedigree, productivity, health, management, and even environmental records such as meteorological data – it’s all connected.
We put it all in a statistical model, and try to make sense out of it. We use current data to predict how we can select and breed the best animals for the future. It’s almost the same as working in the stock market – using historical data to predict the future.
There is a famous idea of ‘if you squeeze the data, it will confess’, and we work hard to draw meaningful inferences from large-scale data sets, then transfer that resource and knowledge to companies, to breeding organisations and to farmers.
I work regularly with breeders, and have begun to interact with farmers. It helps me understand what they need.
What are the main challenges in your field of expertise?
The big challenges are two sides of the same story.
In the Global North, there is limited genetic diversity in livestock populations, which are heavily selected for certain traits, particularly high production. It is hard to adapt these to future challenges such as disease outbreaks or climate change, so various, expensive, management practices need to be implemented to compensate.
Data sets in these populations gather so much data that the models cannot chew this data easily – we need to optimise the models. I am creating algorithms to exploit limited dimensionality of the data, in a way removing aspects that are redundant.
I began this during my PhD, to allow mega-scale evaluations. The algorithm I worked on is now implemented in some of the biggest livestock breeding companies.
On the other hand, in the Global South we have many locally adapted livestock species and breeds, which have some unique genomic properties and are very diverse, but do not have extensive data sets.
I’m trying to understand if we can find the best combination of genes from animals in the South and North, to create resilient cross-breeds or multi-breeds. One of my objectives is to enhance the current livestock breeding strategy by increasing productivity, adaptability and resilience in novel cross-breeding models.
Congratulations on your Roslin Institute Career Track Fellowship. What do you hope to focus on?
For my fellowship, I am developing statistical models for the cross-breeding described earlier. I’m already working with the University of Zagreb and the Croatian Agency for Agriculture and Food to apply such models to Pag and Istrian sheep breeds, which are well adapted to those climates.
I hope to extend the research to other Mediterranean sheep breeds with the University of Sassari, Sardinia, Italy, and to get data from Scottish sheep, to see what is happening at the genome level across breeds, and I am collaborating with a big breeding company from the US. I am hiring a PhD student to work on this.
It is an early stage in my fellowship and my ambitions are to connect genetic variability with climate issues and statistical knowledge to work in the mega-scale genomics space.
What do you enjoy most about your job?
Small victories make me happy. For example, maybe I will be developing a project for months and it isn’t working. Then suddenly, in the middle of the night, I realise there is a bug in my computer code, and I get a result.
Another thing is I am constantly learning; the environment is exciting.
Most of the time, I work alone at my computer and have to really concentrate, but if I get stuck I will share this with colleagues who are more experienced. The animal genetics community is quite open, people share problems and code, and like to help. Open access publishing is very helpful and I strongly believe science should be open.
What challenges do you face?
It’s easy to get frustrated when you’re working hard and not getting the results that you should, but it comes down to how persistent you are. Sometimes in science this is more important than anything else. I always have a slight fear that I’ve made a mistake somewhere in my code, there is always a slight doubt, and this is something I am working to overcome.
On the other hand, however, I think academia is like football – there are limited clubs in the top tier with everyone competing for funding. Not everyone can play in the top tier, and that represents an instability for young scientists, and many decide to switch careers or go to work in the industry.
Submitting grant applications takes huge resources and energy and there is a limit to how much you can influence the outcome. I think the grant awarding system has to be changed. I’m not exactly sure how, but there has to be a fairer way.
What would you like to achieve in your career?
I would like to make a big advance in understanding multi-breed and cross-breed genomics. People have been working on this for 100 years, and there are so many models and theories, but it is not solved completely. I would like to be the guy who figures that out.
The promise of genomics was that it would create the perfect animal ... but of course this is not realistic. There are thousands of genes interacting in organisms, it’s incredibly complex. I’d like to help move that along.
If you could have tea with anyone, who would you choose to meet?
I would choose Steve Wozniak, the technology pioneer and co-founder of Apple. He was with Apple from the start and he stayed out of the limelight, he decided never to sell himself. I’d like to understand him – but people say you should never meet your idols.
If you weren’t a data scientist, what else would you like to be doing?
I’d like to be a programmer or writer for video games. You can use your skills to develop your own world, in a creative, unlimited way. Otherwise, I’d like to play a guitar in a really bad band from Croatia named Mrkve, which means ‘carrots’, just for the fun of it. Thinking about it, I’d like to have tea with Bruce Springsteen – but maybe I should just leave him up on his pedestal.