In the long human struggle to understand why some bodies turn against their own nervous systems, a machine learning tool called RefMap has quietly redrawn the map. Researchers at the University of Sheffield and Stanford University have expanded the known genetic risk factors for motor neurone disease from 15 to 690 — a forty-six-fold leap that transforms a sparse outline into something approaching a full portrait. The discovery does not yet offer a cure, but it opens hundreds of new doors for drug development, genetic counseling, and the broader dream of medicine tailored to the individual. Wh
Machine learning model discovers 690 genetic risk factors for motor neurone disease
690 genes where there were 15 before
Mark
So they went from 15 known risk genes to 690. That's a huge jump. But what does that actually mean for someone diagnosed with motor neurone disease right now?
Mimi
Right now, it doesn't change their treatment. But it changes what's possible. Each of those 690 genes is a potential drug target. Before, researchers had 15 places to look. Now they have 690. It's like going from a small library to a massive one.
Luke
But I want to be careful here. They identified 690 genes using this model. How many of those have actually been validated? The KANK1 gene showed neurotoxicity in lab neurons, which is promising, but that's one gene out of 690.
Mimi
That's fair. KANK1 is the proof of concept—it shows the model is finding real biological signals, not just statistical noise. But you're right that the rest need validation.
Mark
The heritability number jumped five-fold. What does that tell us?
Mimi
It means genetics explains much more of motor neurone disease than we thought. Before, the 15 known genes only accounted for a certain portion of why people get sick. Now the 690 genes account for five times as much of that genetic contribution. There's still more to find, but we're seeing the fuller picture.
Luke
Though heritability is tricky. It's a population-level measure. It doesn't tell you how much any individual person's risk is determined by genes versus environment or luck.
Mark
The tool is generic—they're applying it to other diseases too. Does that make it more or less reliable for motor neurone disease specifically?
Mimi
It's actually a good sign. If the same approach works across multiple diseases, it suggests the method is sound. It's not overfitted to one condition.
Luke
But it also means we don't yet know how well it will perform on diseases with different genetic architecture. Motor neurone disease might have a very different genetic landscape than, say, diabetes or Alzheimer's.
Mark
What's the timeline before any of this becomes a treatment?
Mimi
That's the hard part. Identifying a gene is step one. Validating it takes years. Then designing a drug around it takes more years. We're talking a decade or more, probably, before any of these 690 genes lead to an actual medicine.
Il Polso
- Motor neurone disease has long resisted understanding — with only 15 known risk genes, researchers were navigating a vast genetic darkness with almost no light.
- RefMap's machine learning model tore that limitation open, identifying 690 risk genes by weaving together genetic and epigenetic data in ways traditional methods could not.
- The sheer scale of the expansion — 46-fold — has sent ripples through drug development, genetic testing, and the basic science of how the disease actually works.
- One newly identified gene, KANK1, has already been validated in lab-grown human neurons, producing the same cellular damage seen in real patients — moving the discovery from statistics toward mechanism.
- The tool is not limited to motor neurone disease; the team is already applying RefMap to other complex conditions, suggesting the impact could extend across the landscape of human illness.
- The path from 690 candidate genes to actual medicines remains long, but the coordinates now exist — and that changes everything about where researchers can aim.
In the long human struggle to understand why some bodies turn against their own nervous systems, a machine learning tool called RefMap has quietly redrawn the map. Researchers at the University of Sheffield and Stanford University have expanded the known genetic risk factors for motor neurone disease from 15 to 690 — a forty-six-fold leap that transforms a sparse outline into something approaching a full portrait. The discovery does not yet offer a cure, but it opens hundreds of new doors for drug development, genetic counseling, and the broader dream of medicine tailored to the individual. What science can now see, it may one day learn to treat.
Researchers at the University of Sheffield and Stanford University have built a machine learning system called RefMap that has fundamentally changed what scientists know about the genetic roots of motor neurone disease. By integrating genetic and epigenetic data — the full picture of how genes are expressed and regulated — the model identifies which variations raise a person's risk of developing the condition.
The results are difficult to overstate. Where scientists previously knew of roughly 15 genes associated with the disease, RefMap has identified 690. That forty-six-fold expansion also brings a five-fold increase in measurable heritability — the share of disease risk traceable to genetics rather than environment or chance. Dr. Johnathan Cooper-Knock of Sheffield's Neuroscience Institute notes that each new gene is a potential drug target, and that families with a history of the disease may one day use this knowledge to understand their own risk.
One gene has already moved from discovery toward validation. KANK1, flagged by RefMap as a risk factor, was tested in human neurons grown in the laboratory and produced patterns of cellular damage that mirror what researchers observe in actual patients. That is the difference between a statistical association and a biological mechanism — and it is the kind of evidence that drug designers need.
Stanford geneticist Sai Zhang emphasizes that RefMap was built to travel. It is a generic system already being applied to other complex diseases beyond motor neurone disease. Department chair Michael Snyder frames the larger ambition: machine learning is uncovering the hidden genetic architecture of human illness, and that knowledge is the foundation on which personalized medicine will eventually be built. The cure remains distant, but the map is now far larger than it was.
Researchers at the University of Sheffield and Stanford University School of Medicine have built a machine learning system that has fundamentally reshaped how scientists understand the genetic architecture of motor neurone disease. The tool, called RefMap, works by integrating genetic and epigenetic data—the full landscape of how genes are expressed and regulated in cells—to identify which genetic variations increase a person's risk of developing the disease.
The scale of what RefMap has already accomplished is striking. Before this work, scientists had identified roughly 15 genes associated with motor neurone disease risk. The new model has identified 690. That is not a marginal improvement. It represents a forty-six-fold expansion of the genetic map, and with it, a five-fold increase in what researchers call heritability—the portion of disease risk that can be traced to genetic variation rather than environment or chance.
Dr. Johnathan Cooper-Knock of Sheffield's Neuroscience Institute frames the practical significance plainly: each newly discovered risk gene is a potential target for drug development. The model does not just add numbers to a list. It opens doors. Families with a history of motor neurone disease could eventually use genetic testing to understand their own risk. Pharmaceutical researchers now have hundreds of new biological pathways to explore rather than dozens.
One gene in particular has already moved from discovery to early validation. Called KANK1, it was identified by RefMap as a risk factor for motor neurone disease. When the team tested it in human neurons grown in the laboratory, the gene produced a pattern of cellular damage—neurotoxicity—that mirrors what researchers observe in the brains of actual patients. This is the kind of evidence that suggests a gene is not just statistically associated with disease but actually involved in causing it. It is the difference between a correlation and a mechanism, and it matters enormously for drug design.
Sai Zhang, an instructor of genetics at Stanford, emphasizes that RefMap is not a tool built only for motor neurone disease. It is a generic system designed to work across different complex diseases. The team is already applying it to other conditions in their laboratory. This suggests the real impact may extend far beyond what has been discovered so far.
Michael Snyder, chair of Stanford's genetics department and corresponding author on the work, articulates the longer vision: machine learning applied to genome analysis is uncovering genes that traditional methods miss—the hidden genetic architecture of human disease. That knowledge, he argues, will eventually enable personalized treatment and intervention tailored to individual genetic profiles rather than one-size-fits-all approaches.
Motor neurone disease remains a devastating condition with no cure. But the landscape of what is knowable about its genetic basis has shifted dramatically in a short time. What happens next depends on whether these 690 newly identified genes translate into actual therapeutic targets, whether drug companies can turn biological insight into medicines, and whether the promise of personalized medicine can move from the laboratory into clinical practice.
Citazioni salienti
Each new risk gene discovered is a potential target for the development of new treatments for MND and could also pave the way for genetic testing for families to work out their risk of disease.— Dr. Johnathan Cooper-Knock, University of Sheffield's Neuroscience Institute
By doing machine learning for genome analysis, we are discovering more hidden genes for human complex diseases such as MND, which will eventually power personalized treatment and intervention.— Michael Snyder, Stanford University School of Medicine
Vuoi la storia completa?
Leggi l'originale su News-Medical.net ↗
Education