For people living with a rare disease, getting an answer can take years. Symptoms may involve several organs, mimic more common conditions or fail to fit neatly into a known diagnosis. Even when genetic testing is performed, finding the disease-causing variant is only part of the challenge: clinicians must determine what that genetic change means in the context of the patient's symptoms.

Now, Chinese researchers have developed an artificial intelligence system designed to help bridge that gap.

Called OneGenome, the open-source system combines a genomic foundation model called Genos with large language model capabilities. The aim is to help interpret DNA sequences alongside medical knowledge and identify genetic changes that may be relevant to rare diseases. According to its developers, the system has outperformed general-purpose AI models and conventional gene models in several diagnostic and medication-guidance tests.

The development is significant because rare disease diagnosis is already moving towards genomic sequencing and computational analysis. But the technology is best understood as a decision-support tool, not an AI doctor capable of independently diagnosing patients.

What is OneGenome and how does it work?

The human genome contains roughly 3 billion DNA base pairs, creating an enormous amount of information for researchers and clinicians to analyse. A genetic test can identify differences, or variants, in that DNA. The difficult part is determining whether a particular variant is harmless, potentially disease-causing or relevant to a patient's specific symptoms.

OneGenome is designed to help with this interpretation. According to reports, the system combines Genos, a human-genome foundation model developed by BGI-Research and Zhejiang Lab, with large language model capabilities. Genos was designed to analyse long genomic sequences and represent genomic information from diverse populations. OneGenome then adds reasoning and medical-language capabilities to help connect genomic findings with clinical information and medical literature.

In July 2026, the system received an Innovate for Impact Use Case Award from the International Telecommunication Union at the UN AI for Good Global Summit for its open-source application in rare-disease decoding. BGI-Research says OneGenome is intended to make advanced genomic AI more accessible for research and precision medicine.

Why is AI useful in rare disease diagnosis?

Rare diseases are often difficult to diagnose precisely because individual conditions are uncommon and their symptoms can overlap with many other disorders.

Genomic sequencing has already changed this landscape. Whole-genome sequencing examines far more of the genome than conventional targeted genetic tests, while whole-exome sequencing focuses on protein-coding regions, which contain a large proportion of known disease-causing variants.

A 2024 study published in Nature Genetics used genome sequencing in people suspected of having rare diseases and identified plausible diagnoses in 29.3% of participants, including many whose diagnoses had previously been missed by other testing approaches.

More recent evidence also suggests that genomic sequencing can produce clinically useful diagnoses in a substantial proportion of people with suspected genetic disorders. A 2025 study involving 1,452 Korean families found a molecular diagnosis in 46.2% of families, with the results influencing clinical management in 18.5% of diagnosed cases.

AI could potentially make this process more efficient by helping clinicians sift through enormous amounts of genomic and medical information.

But OneGenome is not simply "AI reading your DNA"

This distinction is important.

DNA does not come with a straightforward label saying "this mutation causes disease". Genetic interpretation requires information about the variant, the patient's symptoms, family history, population frequency, existing research and the strength of available evidence.

OneGenome is intended to connect these different layers of information. Rather than simply matching a DNA sequence to a database entry, the system is designed to use reasoning to interpret the possible clinical consequences of genetic variants.

That approach reflects a broader shift in medical AI: from systems that simply recognise patterns towards systems that can integrate multiple sources of evidence. Another Chinese AI system, DeepRare, developed by researchers at Shanghai Jiao Tong University and Xinhua Hospital, illustrates the same direction. Published in Nature, DeepRare combines clinical phenotypes, medical literature and genomic information to generate ranked rare-disease possibilities with traceable evidence.

In testing, DeepRare achieved a top-ranked diagnostic rate of 57.18% using phenotype information alone, while performance exceeded 70% when genetic information was incorporated, according to the research team's reported results.

Could this reduce the "diagnostic odyssey"?

Potentially, yes, but that remains to be demonstrated in routine healthcare. People with rare diseases can undergo repeated consultations, scans and genetic tests before receiving a molecular diagnosis. The NIH's Undiagnosed Diseases Program, for example, uses multiple genomic sequencing technologies to investigate rare and previously unexplained conditions. 

AI systems could help clinicians prioritise which genetic variants and disease possibilities deserve attention, potentially reducing the amount of time spent manually searching through huge datasets. For countries where access to specialist genetic expertise is limited, that could be particularly valuable.

However, accuracy in a research benchmark is not the same as proven clinical effectiveness. OneGenome's developers' reported performance needs to be independently evaluated across different populations, healthcare systems and real-world patient groups.

What are the limitations?

Genomic AI faces several challenges.

  • First, a genetic variant may have uncertain significance. Finding a variant does not automatically prove that it caused a patient's illness.
  • Second, genomic datasets can be unevenly representative. Genetic diversity matters because a variant that is rare in one population may be relatively common in another. This is particularly relevant for countries such as India, with its enormous population and genetic diversity.
  • Third, privacy is a major consideration. Genomic information is uniquely sensitive because it can reveal information about an individual and, potentially, biological relatives.
  • Finally, AI-generated results require clinical interpretation. Genetic testing is generally most useful when combined with a patient's phenotype, family history and appropriate genetic counselling.

What does this mean for patients?

The emergence of tools such as OneGenome does not mean people should upload their DNA to an AI platform and expect a diagnosis. Instead, the technology represents another layer in the evolving field of precision medicine.

For someone with an unexplained condition, genomic sequencing may sometimes help identify an underlying genetic cause. AI could increasingly assist specialists in interpreting those results faster and connecting them with the expanding medical literature. But the final diagnosis still needs to be established through appropriate clinical evaluation and, where necessary, confirmatory genetic testing and specialist review.

China's OneGenome is part of a rapidly developing intersection between artificial intelligence and genomics. By combining a human-genome foundation model with large language model capabilities, the open-source system aims to make the interpretation of complex genetic information more accessible and useful for rare-disease research and diagnosis. Its promise lies not in replacing doctors, but in helping them navigate information that is simply too vast for humans to process manually.

For people caught in years-long searches for a diagnosis, even making that search faster could eventually make a meaningful difference. But before such systems become routine clinical tools, they will need rigorous validation, safeguards for genomic privacy and careful oversight by medical professionals.



Disclaimer: This content including advice provides generic information only. It is in no way a substitute for a qualified medical opinion. Always consult a specialist or your own doctor for more information. NDTV does not claim responsibility for this information.