Artificial intelligence language models have successfully identified the genetic causes of rare diseases and hearing loss without any medical training or fine-tuning, according to a new study from Stanford University researchers working with Google DeepMind. The breakthrough suggests that AI could dramatically accelerate genetic diagnosis for an estimated 350 million people globally living with undiagnosed genetic conditions.

How Do AI Models Identify Disease-Causing Genes?

The Stanford team developed an AI pipeline that processes lists of candidate genes and ranks them by likelihood of causing disease. The system works by asking the AI model to research genes against medical literature, cross-reference them with a patient's specific symptoms, and then explain its reasoning with cited evidence. This approach bypasses the traditional bottleneck in genetic research: the need for human geneticists to manually sort through thousands of genetic variants to separate true disease-causing mutations from false positives.

The researchers tested two different large language models, or LLMs (AI systems trained on vast amounts of text to understand and generate language). Med-PaLM 2 is specifically trained on medical information, while Gemini 2.5 Pro is a general-purpose reasoning model with no medical fine-tuning. Remarkably, both successfully identified disease-causing genes, with Gemini 2.5 Pro performing just as well despite having no specialized medical training.

What Were the Study Results?

In the first phase, Med-PaLM 2 analyzed genes from mouse studies and not only correctly identified known disease-causing genes but also discovered a new genetic factor responsible for spontaneous hearing loss. This finding was then validated experimentally in the laboratory. The team then moved to human studies, testing the AI pipeline on 20 patients with hearing loss and 6 patients with rare genetic diseases.

Gemini 2.5 Pro successfully identified genetic factors causing hearing loss in all tested patients, matching the diagnoses made by clinical geneticists and ear specialists. For the more complex rare disease cases, researchers modified the pipeline to account for multiple overlapping symptoms, and the AI again successfully pinpointed causative genetic variants. The results were published in the journal Advanced Science.

The implications are significant. Hearing loss alone affects one-third of adults over age 61 and 80 percent of those over 85, with roughly half having a genetic cause. Yet many genetic factors remain unidentified, and the significance of genetic variants in hearing loss patients is largely unknown.

Why This Breakthrough Matters for Genetic Medicine

• Speed and Cost Reduction: Current methods for discovering genetic disease factors rely on genome-wide association studies (GWAS), which are expensive, time-consuming, and require expert clinical geneticists. AI-based analysis could dramatically reduce both cost and turnaround time.
• Access to Precision Medicine: AI-enabled genetic diagnosis could provide access to precision genomic health for billions of individuals who currently lack it, particularly in regions with fewer genetic specialists.
• Faster Treatment Development: For the estimated 350 million people worldwide with rare genetic diseases, identifying the responsible genes could expedite the development of targeted treatments and therapies.

"We demonstrated that large language models could facilitate genetic discovery in mice and could generate genetic diagnoses for hearing loss and rare genetic diseases in humans," said Gary Peltz, Professor of Anaesthesiology, Perioperative and Pain Medicine at Stanford University School of Medicine.

Gary Peltz, Professor of Anaesthesiology, Perioperative and Pain Medicine at Stanford University School of Medicine

The current standard approach, GWAS, identifies statistical associations between genetic variants and disease traits. However, this method generates false positives alongside true disease-causing variants, requiring expert interpretation to separate signal from noise. The AI pipeline essentially automates and accelerates this expert reasoning process.

What's Next for AI-Powered Genetic Discovery?

The Stanford team is already planning the next phase of development. They are integrating the AI pipeline into agentic frameworks, which would enable autonomous AI agents to work with plug-ins containing data about gene mutations and proteins. This could further improve the pipeline's accuracy and capabilities.

Looking further ahead, researchers hope to develop models capable of reading patients' electronic medical records and analyzing their genomes in a fully automated way. This would represent a shift from disease treatment toward disease prevention, with customized prevention plans developed based on each person's genetic risk factors.

"These advanced LLMs will have a dramatic impact on biomedical research. They represent a new type of knowledge-generating AI that has reasoning and improved agentic capabilities. This will enable those LLMs to produce novel hypotheses that can catalyze biomedical discoveries, accelerate finding new treatments, and improve healthcare," noted Peltz.

Gary Peltz, Professor of Anaesthesiology, Perioperative and Pain Medicine at Stanford University School of Medicine

However, experts emphasize that human oversight remains essential. Clinicians and geneticists will continue to play a critical role in evaluating and interpreting AI output. Additionally, a major challenge ahead involves analyzing the 98 percent of the genome that doesn't code for proteins. Most current genetic analysis focuses on the roughly 2 percent of DNA that encodes messenger RNA and proteins, leaving vast stretches of genetic information poorly understood.

The Stanford research represents a significant step toward making genetic diagnosis faster, cheaper, and more accessible globally. By demonstrating that general-purpose AI models can reason through complex genetic data without specialized medical training, the study opens new possibilities for democratizing precision medicine and helping millions of people find answers to their genetic health questions.