Most of the traits we care about in agriculture are not single gene traits, observes Avalo CSO Mariano Alvaro, PhD. They might result from hundreds, or even thousands of genes working together in ways we don’t understand. So how can breeders cut through all the biological noise to identify what’s important?
At North Carolina-based Avalo, which deploys interpretable machine learning (IML) to develop more climate resilient varieties of crops such as sugarcane and cotton, the key is embracing complexity, he says. “We look at the forest, not the trees.”
Through its rapid-evolution platform, Avalo taps into broader gene pools, wild varieties, and natural pollination to generate richer datasets; uses computer-based modeling to spot promising genetic traits more quickly across many samples; and applies machine learning to make sense of complex genomes “to find underlying characteristics, not just individual genes.”
In a notoriously challenging crop such as sugarcane, which has a huge, complex genome and long breeding cycles, bringing new varieties to market can take well over a decade, he says. Avalo—which is working with Coca-Cola Europacific Partners (CCEP) to develop more climate-resilient varieties in Australia—reckons it could potentially cut this timeline in half.
AgFunderNews (AFN) caught up with Alvarez (MA) at the recent World Agri-Tech event in London to find out more…
AFN: How have breeders traditionally connected genes with traits of interest?
MA: It’s actually a hugely challenging process, even though this is something that we’ve been doing for decades. At this point, there’s really two approaches, forward and reverse. So either you are going in, gene by gene, breaking something and then testing the plant to see how it responded, either it did or it didn’t affect the trait that you’re interested in.
Or you can challenge the plant with a particular condition and then measure the expression of genes and try to discover it that way.
Both of those methods can be very challenging.
In the first case, it takes a lot of time and effort and money to create very precise gene edits that you really want if you are looking to discover the genes for a complex trait, while the associative methods just have a lot of noise in them. Biology is complicated; there are many, many genes responding to each other and to their environment, so it can be difficult to disentangle that.
In a normal discovery process, thousands of genes [might be] responsible for any given trait, so, I think the days of easy discovery are over.
AFN: How does Avalo’s process work?
MA: We use associative methods, but we build on them using machine learning techniques. Rather than looking at each individual gene one at a time, we use machine learning to look at the entire genome all at the same time, look at the response of plants in different environments, and then identify which genes are responsible.
And of course, we still see a huge number of genes that are responsible, but we can narrow that down much more than is possible today and eliminate a lot of that sort of biological noise and just focus on a constellation of key targets for the next parts of our process.
AFN: How does your predictive modeling platform work?
MA: When we discover the genetic basis of a complex trait, we then turn that into a predictive algorithm. So imagine if you were navigating by the stars, you wouldn’t look at every star in the sky. You would focus on a few specific constellations to guide you. And that’s basically what we’re doing.
We design assays that are only looking at the genes we are interested in and that allows us to make much more predictive models, because those models aren’t confounded by the sort of underlying biological noise.
AFN: What is your business model?
MA: We really think about ourselves as an integrated crop development company, which means we actually run our own breeding programs.
We use our models in our own programs, and we use them to lower the cost of development. Yeah, type of development. We then work with growers to get them our seeds and provide them with sets of agronomic recommendations that go along with those seeds.
And then, in some cases, we actually help then sell the final product post farm gate. In the case of sugarcane, we’re working with Coca-Cola Europacific Partners (CCEP). They source all of the sugar in their supply chain in the Pacific from sugarcane coming from Queensland, Australia.
AFN: How challenging is sugarcane to work with? And what are you specifically looking for with this partnership? Is it about growing it in places with less water? Is it about using less fertilizer?
MA: Sugarcane is a hugely challenging crop to work on. It has an incredibly complex genome, many times more complex than the human genome. It’s a perennial crop, and so it has a very long breeding cycle. But that is exactly the type of project we want to work on at Avalo, because it has such an impact on the supply chain.
For CCEP, it’s 16% of their total Scope three emissions, which is an enormous number. And so if we’re able to make a difference here, we can actually have a huge impact from a sustainability perspective.
AFN: What’s the status of your work with Coca-Cola Europacific Partners?
MA: So this year is really about learning for us, going into the supply chain, working with a small core set of growers, and figuring out how we can use our technology, our models, and eventually our genetics, to make a difference in the system.
AFN: What other crops are you looking at?
MA: A huge amount of our work and our resources go into cotton. We’re growing cotton primarily up in the panhandle of Texas, and we’re working with other elements of the supply chain, textile mills, brands, to be able to get that product into the market.
This year, we’re producing on 2,000 acres in Texas. Next year, we hope to more than double that.
Our broccoli work is in the process of being commercialized with a partner in New York. Our rubber work is ongoing. Unfortunately, that is work with the government, so I can’t talk about it, but it’s really exciting work looking at domestic sources of rubber.
In rice, we have a lot of expertise. We’ve done grant funded work over the course of our development as a company. But an important thing for us at Avalo is that our scale up is really led by elements in the industry and supply chain that see a really burning need. We’re scaling sugar and cotton because we have really good commercial partners.
AFN: After you’ve identified genes associated with traits of interest would the next step be gene editing, genetic engineering or traditional plant breeding?
MA: We’re really focused on traditional breeding because we think it has the lowest cost profile, a shorter regulatory process. That means we can get a product out to market faster and spend less money on R&D.
AFN: Is it unusual to have a CPG company like Coca Cola getting this involved right up the supply chain for a key ingredient?
MA: CCEP has been an incredibly supportive partner. And I think a lot of their interest reflects change in the industry as a whole, where if you want to be able to tackle supply chain issues, Scope three emissions or quality issues, you have to be able to look all the way back through the supply chain.
Coca Cola has invested in our company and is also supporting some of the R&D process, which is really helpful for us as a startup, because it gives us time to understand how to build a sustainable and scalable business model in this space that ultimately is supported not just by Coca Cola, but by the mills and the growers and the entire supply chain.
👉 Watch more videos from World Agri-Tech and Future Food-Tech
The post 🎥 Avalo harnesses ‘whole genome AI’ to accelerate crop breeding: ‘We look at the forest, not the trees’ appeared first on AgFunderNews.
