In the world’s breadbaskets — from the wheat fields of the Canadian prairies to the soybean plots of the American Midwest — a silent revolution is unfolding. But it’s not happening in the fields. It’s taking place in the unseen molecular circuitry of plants, decoded not by farmers or even traditional scientists, but by artificial intelligence.
Welcome to the new frontier of agriculture, where the blueprint for tomorrow’s crops lies not in how we grow them, but in how well we can read — and transform — their DNA.
At the forefront of this revolution is Christian Kubica, bioinformatics scientist at Computomics, a biotech firm based in the university town of Tübingen, Germany. Unlike the bustling laboratories that conjure images of test tubes and microscopes, Computomics operates more like a command center of code and computation, an AI powerhouse that helps breeders do something that, until recently, seemed nearly impossible: find the genes that truly matter.
To Kubica, the phrase “gene discovery” has always felt a little misleading.
“When people hear that term, they often think of gene annotation: figuring out which parts of the genome code for proteins,” he explains. “But that’s not what we’re doing. We’re not just labeling the parts. We’re pinpointing which genes actually do something important — like confer disease resistance or help a plant survive drought — and identifying the genetic variation that controls those traits.”
That’s where pan-genomes come in, a powerful alternative to the single-reference genomes of the past. Unlike the traditional linear genome, a pan-genome captures the full range of genetic diversity across multiple individuals within a species. It’s not a straight line; it’s a network, a graph, a dynamic map of roads that evolution could have taken.
“With a linear genome, you only get one version of the truth,” Kubica says. “But with a pan-genome, we can see all the possibilities across the individuals present in the pangenome, and that gives us the power to spot meaningful variation that was invisible before.”
Traditionally, breeders searching for a gene associated with a trait, such as resistance to a crop disease, relied on genotyping methods that provided just specific genetic markers like single nucleotide polymorphisms (SNPs), which are restricted in their ability to capture the full spectrum of genetic variation. As Kubica notes, such markers are frequently too small to fully explain genetic function, potentially overlooking critical variations that influence traits.
In contrast, modern genotyping methods have significantly expanded the capacity to detect structural genetic variation. This broader perspective enhances the understanding of genetic influences on traits, offering more precise and insightful information than traditional methods.
“A single base change doesn’t necessarily alter the function of a protein in a meaningful way,” he says. “But larger variations like deletions, duplications, or structural rearrangements within a gene almost always have a profound impact. And those are exactly the kinds of variants pan-genomes allow us to map.”
And unlike many groups still stuck analyzing SNPs, Computomics goes several steps further: using AI to associate complex genetic structures with phenotypes, and even uncovering variants located outside of genes, such as in regulatory regions, areas long overlooked but crucial in determining how genes behave.
The result? Breeders aren’t just handed a marker near a gene of interest. They get the actual causal variant, the precise change responsible for a trait.
Kubica shares one example of how Computomics AI-powered technology cracked a genetic mystery.
A customer had spent years trying to understand why certain lines in their breeding program resisted a specific pathogen. Traditional analysis had come up empty. But when they used Pantograph, a platform built to investigate deep pan-genome graphs of public and proprietary data, they hit gold.
“They were able to genotype their own lines against the pan-genome and actually found the causal gene and variant responsible for resistance,” Kubica says. “It had been hidden in structural variation. They told us: ‘Without Pantograph’s visualization of the pan-genome, we never would have found it.’ This single discovery alone has already yielded a positive return on their investment into Pantograph.”
In another project, Computomics is now applying genome-wide association studies (GWAS) directly to the pan-genome graph — linking traits to full DNA sequences rather than to isolated SNPs. “It’s a whole new way of running GWAS,” Kubica says. “We’re going from correlation to causation.”
In the face of climate change, legislative shifts, and rising demand for food, this type of precision isn’t just nice to have — it’s a necessity.
“As regulatory frameworks change, especially in places like the European Union, breeders will need to know the exact variant controlling a trait,” Kubica explains. “Approximate markers won’t be enough anymore. You’ll need to show exactly what you’ve modified and why.”
That demand is driving Computomics’ next phase: empowering breeders to run these analyses themselves. A future update to Pantograph will allow users to perform pan-genome GWAS on their own systems, keeping their data private while leveraging Computomic full AI toolkit.
It’s a reflection of the company’s philosophy: cutting-edge, but customizable. “Most of the new features we build come directly from customer requests,” Kubica says. “Our roadmap evolves based on what breeders actually need.”
Kubica’s own journey mirrors the evolution of this field. His mother was a human geneticist. His father, a gardener. He ended up in plant bioinformatics — a marriage of both worlds. With a PhD in hand and a love of biology coded into his DNA, he embodies the blend of field and lab that the future of agriculture demands.
“You can’t solve these problems with code alone,” he says. “You need to understand how plants grow, how environments change, how traits behave in the real world.”
That real-world complexity is what makes Computomics’ AI so powerful — and what makes gene discovery, redefined through the lens of pan-genomes and machine learning, more than just a buzzword.
It’s a new language for decoding life itself.
And yet, for all the computing power and genomics wizardry, Kubica is quick to emphasize one thing: technology won’t replace breeders.
“These tools don’t make decisions,” he says. “People do. What we offer is clarity — a clearer signal in the noise. It’s about empowering human intuition with deeper insight.”
Originally posted on 21st August, 2025 on SeedWorld US
Share on