Plant breeding was at one time as much art as science. Plant breeders would cross lines they felt had genetic potential to produce new varieties of progeny with the hope of boosting yield, or adding disease resistance. The process was not unlike what farming ancestors did when they selected more desirable-looking plants to save and replant the following year. It was that process used to change early teosinte grass plants into what would eventually become corn.
And until the last 20 years, the process was pretty much unchanged. The characterization of parent lines improved, which also offered enhanced yield potential. Yet in the last two decades there has been a step change in the breeding process thanks to the use of genetic markers. Add in advanced data analytics and the process improves further.
Tracy Doubler, head of North American breeding projects, Syngenta, outlined the changes his company has undertaken in its move to boost soybean performance. He explains that one of the first innovations to come to soybean breeding was mechanization for harvesting plots, which allowed all companies to put in more plots and harvest more for evaluation.
"But to identify a high-yield soybean variety that would sell 1 million units, that was like searching for a needle in a haystack," he says. "You have environmental error and you can't sample all of the environments during a research and development trial."
Essentially, the ability to put soybeans through all potential environments for testing is a challenge. He explains that in the 1970s companies could harvest 300 yield trial plots a day in two 12-foot rows. But to get good information, they would need to test thousands of lines in all environments for growers and in multiple replications to find the results. They were limited by mechanization. That 300 yield plots a day just wasn't cutting it.
In the 1980s, the equipment got better, and they could harvest 1,500 plots per day.
And today? Mechanization has advanced so that Syngenta can harvest 3,000 trial plots per day and that combine captures yield data for each at the same time. That's a lot of information and capability but how do you sort that to determine the best out of all those trials? And separate yourself in the market, all other seed companies are using the same mechanical tech.
Doubler says that molecule markers added to breeding programs made a difference. He explains that the breeding process is like a funnel. "It's easy to create genetic diversity to find high-yield, advanced progeny, but no one can test every line," he says. "There are millions of varieties and you cannot test them all."
The key would be to develop a system that would stack the deck in favor of high-yield lines but yield is the result of many factors, there isn't a single gene related to yield.
One improvement of molecule markers was the reduction in the cost to search for genes. While it might cost $10 to $15 to phenotype a specific soybean line, finding traits with molecular markers cost 5 to 10 cents. Discovering a specific resistance gene or other traits became more cost effective. But now you're back to millions of lines to search out, to screen for defensive traits.
With that tech, companies are not just using molecular markers but they're working on ways to predict phenotype performance for complex traits - iron chlorosis, sudden death syndrome and ultimately yield, Doubler says. How to raise bar? Data analytics.
"Data analytics is differentiating us in the market," Doubler says. "We are using modeling, simulation, optimization and operations research combined with advanced math to optimize the breeding program."
When confronted with the molecular-marker pile of data plant breeder, data analytics was the way to go. He shows a field where yields vary in the field. In the old model, a breeder might pick the lines that yielded best in the trail and move forward with those. However, a review of molecular markers in the varieties, combined with analytics across many trials may show that in one field the lower-yielder may be the better performer.
"We're using data-driven decisions in every step of the plant breeding process - from testing locations, to tools for optimum trait introgression," he explains. The company even uses the information to determine the optimal size of trials for a plant breeding program to be successful.
Doubler says he believes Syngenta is the first, and perhaps only, company to bring data analytics to plant breeding on this level. And the company is being recognized for the work. In 2015 it won the Franz Edelman Award, which recognizes company use of operations research - a prestigious award that in the past has been given to such companies as the Centers for Disease Control and Hewlett Packard.
And for 2016, Syngenta was awarded the Association of National Advertisers Genius Award in Analytics Innovation for its creative analytical approach taken to solve complex problems in plant breeding. Other honorees this year include The Clorox Company, Hilton Worldwide and Turner.
Data analytics is high-end math that can sort out the bad information from a breeding trial, say an irrelevant environmental issue, or a mechanical problem from the harvesting combine. Doubler confides that through the company's plot work, they determined that the latest model plot harvester actually had a data collection flaw that was later rectified.
Making a difference
When farmers hear Big Data they often think it applies to their farms, and in many ways it does. But the very suppliers you buy from are maximizing their work through the use of high-end data analysis in new ways. Syngenta is being recognized for its work, next step is for that work to show up in higher-yielding soybeans.
"We started incorporating analytics in 2010 to support breeding, and aggressively measure the rate of gain. Across all maturities we support an additional 2 bushel per acre per year gain, well above the industry standard," Doubler says.
He says that this rate of gain would allow farmers to double the amount of soybeans produced in the world by 2050.
The work is also helping the company keep up with an ever-changing biotech challenge - getting new traits into top yielding varieties without taking a yield hit. "We sell no conventional soybeans in the U.S.," Doubler says. "But we maintain a conventional breeding program focused on higher yields and defensive traits."
When a new trait comes along, Syngenta is bringing that trait into a high-yielding variety, and their program allows them to shorten time to market with top-yielders and new traits. "We were years behind others in getting the Roundup Ready 2 Xtend trait for soybeans," Doubler says. "But what I can tell is that we'll have products on the market in the same time frame as Monsanto, and in the same timeline we'll have more maturities."
This shows how this data analytics combined with plant breeding can be used in a number of ways - to boost yield, and to speed traits to market.
The plant breeding world has come a long way.