New AI platform identifies disease biomarkers with over 90% accuracy

In a new study published in Nature Medicine, researchers analyzed stool, blood, and other health data from nearly 250 people using a custom-built AI platform.

People suffering from the mysterious and often misunderstood condition known as chronic fatigue syndrome may finally be closer to getting a clear diagnosis—and perhaps, in time, targeted treatments. Scientists have used artificial intelligence to identify patterns of disruption in gut microbes, immune cells, and metabolism that together signal the disease with surprising accuracy.

In a new study published in Nature Medicine, researchers analyzed stool, blood, and other health data from nearly 250 people using a custom-built AI platform. The tool—called BioMapAI—accurately distinguished patients with myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) from healthy individuals about 90% of the time.

“Our study achieved 90% accuracy in distinguishing individuals with chronic fatigue syndrome, which is significant because doctors currently lack reliable biomarkers for diagnosis,” said study author Dr. Derya Unutmaz, an immunologist at The Jackson Laboratory.

Doctors have long struggled to define or even confirm the existence of ME/CFS, partly because its symptoms vary so widely—from fatigue and brain fog to dizziness and chronic pain—and partly because no simple blood test can detect it. Some physicians have even dismissed the illness as psychological.

The new research suggests that biological signatures of the disease are real and measurable. Led by microbiologist Julia Oh, formerly at Jackson and now at Duke University, the team used multi-omics data—microbiome sequences, immune cell profiles, blood chemistry, and clinical symptoms—to chart disruptions in 12 key symptom categories, including sleep disturbances, gastrointestinal issues, and cognitive impairment.

“We integrated clinical symptoms with cutting-edge omics technologies to identify new biomarkers of ME/CFS,” Oh said. “Linking symptoms at this level is crucial, because ME/CFS is highly variable.”

The analysis found that ME/CFS patients had lower levels of helpful fatty acids like butyrate in the gut, and higher levels of potentially problematic compounds like benzoate and tryptophan. These metabolic changes were linked to an overactive immune response, including imbalances in MAIT cells—a type of immune cell that monitors microbial activity in the gut.

“MAIT cells bridge gut health to broader immune functions, and their disruption alongside butyrate and tryptophan pathways, normally anti-inflammatory, suggests a profound imbalance,” said Unutmaz.

The study also found that patients who had been ill for more than a decade showed more severe disruptions than those with shorter disease durations. That finding could help explain why symptoms often become more entrenched over time—and why early diagnosis might make a difference.

“Our data indicate these biological disruptions become more entrenched over time,” Unutmaz said. “That doesn’t mean longer-duration ME/CFS can’t be reversed, but it may be more challenging.”

The researchers used data from 153 ME/CFS patients and 96 healthy controls provided by the Bateman Horne Center, a clinic specializing in chronic fatigue, long COVID, and fibromyalgia. The BioMapAI model performed similarly well when tested on outside data, giving scientists confidence that the biomarkers it identified are consistent across patient groups.

“Despite diverse data collection methods, common disease signatures emerged in fatty acids, immune markers, and metabolites,” Oh said. “That tells us this is not random. This is real biological dysregulation.”

Although the findings don’t yet translate into a clinical test or treatment, they mark an important step toward both. Because the microbiome and metabolome can potentially be changed through diet, lifestyle, or drugs, researchers believe they may offer more accessible targets for future therapies than genes alone.

“The microbiome and metabolome are dynamic,” Oh said. “That means we may be able to intervene—through diet, lifestyle, or targeted therapies—in ways that genomic data alone can’t offer.”

The team plans to share their data publicly and continue developing BioMapAI to map other diseases with overlapping symptoms. The hope is to build a more precise understanding of how immune function, gut microbes, and chemical signals interact—potentially offering answers not only for ME/CFS, but for long COVID and related chronic illnesses as well.

“Our goal is to build a detailed map of how the immune system interacts with gut bacteria and the chemicals they produce,” Oh said. “By connecting these dots we can start to understand what’s driving the disease and pave the way for genuinely precise medicine that has long been out of reach.” 

The study has been published in Nature Medicine.

Sanket Mungase
Sanket Mungase
Sanket Mungase is a freelance science writer who covers everything from science, space, robotics, and technologies that change our world. He holds a degree in Mechanical Engineering.