Fine-tuning the corn genome: High-performance computers help eliminate bad mutations

When you’re trying to feed 10 billion people, minimizing the mutations that reduce crop yield ranks as a pretty high priority. By harnessing the power of Cornell high performance computers, scientists have discovered a way to predict where unwanted mutations are likely to occur.

Think of DNA as a car ride through time, says Eli Rodgers-Melnick, postdoctoral associate in the USDA-Agricultural Research Service Buckler Lab for Maize Genetics and Diversity at Cornell University, US. In an article recently published in Proceedings of the National Academy of Sciences, he describes how a genome barrels through the ages. Though mutations sometimes contribute to an adaptive advantage (say, adding a more fuel efficient carburetor), many other mutations bring undesirable characteristics (like a trailer with a flat tire, decreasing the car's fuel efficiency).

In his example, the fuel-efficient car represents the mutation breeders seek to save for future generations. Unfortunately, without a process to separate the good mutations from the bad, the rusty trailer comes along for the ride. Rodgers-Melnick and his team used DNA sequenced from 7,000 corn plants to find the spots where the genome breaks when the plant sex cells are formed. They found that the rate of this breakage process — called recombination – varies predictably across the corn genome.  

Rodgers-Melnick and co-lead author Peter Bradbury set out to see if they could use recombination to predict where bad mutations were located in the corn genome. Mathematical biologists have long predicted bad mutations gather where recombination occurs very rarely. Now that DNA sequencing has become more economically accessible, geneticists can test the theory. In Rodgers-Melnick’s study, they found that bad mutations in corn were preserved significantly more often in low recombination portions of the genome.

“While our tools are still in the early stage, this shows we can spot bad mutations directly from a DNA sequence. With improvements, we will be able to spot with accuracy, and then remove them from the genome by breeding or editing of the genome,” says Edward Buckler, senior author of the study. “Nearly exactly the same tools can be applied to any crop species.”

Management of such a large genomic analysis was only possible with access to Cornell BioHPC labs resources, Rodgers-Melnick notes. Cornell cloud computing capabilities enabled him to run Markov Chain Monte Carlo style algorithms to infer where extremely high rates of recombination occurred. High-performance computing resources made it possible to align and impute approximately 1 million genetic markers across 7,000 corn lines, both remotely and in parallel.

The upshot: “These bad mutations won't be easily removed using conventional breeding techniques. We will need to rethink how we do plant breeding if the rate of grain yield gains achieved in the 20^th century is to be sustained throughout the 21^st and meet the demands of a growing population in a rapidly shifting climate,” Rodgers-Melnick says.

Fortunately, genomic editing technologies are quickly maturing, so the researchers’ vision of future plant breeding is becoming a reality. Rodgers-Melnick expects long-term effects of genome editing to follow the yield improvements in hybrids cultivated through the 20^th century. Scientists assume this increased vigor is a product of complementing mutant genes with good copies so that desirable phenotypes are expressed instead. 

Despite the promise posed by his research, Rodgers-Melnick knows genetic engineering faces resistance. “I realize there is quite a bit of misunderstanding concerning genetic modification. However, we're not interested in taking a gene from some foreign entity and introducing it into corn. Rather, we'd like to go in and fix the bad mutations already present in the natural populations — essentially, genomic medicine for plants.”