A new framework called "near-perfect genome sequencing" combines long-read DNA analysis, artificial intelligence, and diverse genetic reference data to diagnose rare diseases that have stumped doctors for years. Instead of requiring patients to undergo multiple expensive tests, this single-test approach could deliver answers faster and more accurately than current methods.

Why Current DNA Tests Leave Crucial Gaps?

The human genome is enormous, and the tools scientists use to read it have significant blind spots. Short-read sequencing, the most common method today, breaks DNA into tiny fragments of about 100 base pairs and reassembles them like a jigsaw puzzle. The problem: many pieces don't map perfectly to the reference genome, leaving gaps that hide disease-causing variants.

Alexander Hoischen, a genomic technologies researcher at Radboud University, explained the limitation this way: "Imagine the human genome being a jigsaw puzzle with a million pieces. The trouble with short reads is, they are tiny pieces." This fragmentation is especially problematic for rare disease diagnosis, where patients often need to test for variants outside known disease-related genes. The result is a diagnostic odyssey: multiple tests, high costs, and still no answers.

Hoischen and his team tested long-read sequencing, which reads DNA fragments up to 20,000 base pairs at a time, on samples from over 800 people with rare diseases. The results were striking: long-read sequencing alone diagnosed 160 patients conclusively, while standard-of-care diagnostic testing, which involved multiple tests, diagnosed only 137 patients.

What Makes "Near-Perfect" Genome Sequencing Different?

Long-read sequencing alone, however, still has limitations. Even longer reads may not be long enough to complete the entire jigsaw puzzle. Rather than adding another expensive sequencing method, Hoischen's team proposed a framework that combines three key components:

• Long-read sequencing foundation: Reads larger DNA fragments to capture more context and reduce gaps in the genome sequence.
• Diploid genome assembly: Represents both copies of each chromosome to provide a complete individual genome picture.
• Diverse pangenome references and AI interpretation: Uses genetic data from people of different ancestries and artificial intelligence tools to identify novel disease-causing variants that traditional methods miss.

The AI component is critical. Even with perfect sequencing, interpreting the data remains a challenge. Hoischen noted: "Even if we sequence the genome to perfection, we need these AI tools and a few other concepts to also make the understanding near perfect".

Hoischen

How Could This Transform Rare Disease Diagnosis?

The implications are significant for patients who have spent years seeking answers. Rare disease diagnosis typically involves a battery of tests, each targeting specific genes or conditions. Many patients remain undiagnosed or undertreated because their variants fall outside the scope of standard testing. A single, comprehensive test could change that.

Hoischen emphasized the human impact: "We would help our doctors so much more efficiently. We could help the patients that deserve clearer answers much more rapidly and much better. So that's a massive hope".

Hoischen

Beyond diagnosis, this approach could also help researchers identify entirely new rare diseases. By examining variants that don't match known disease patterns, scientists may discover disease-causing mutations that have never been documented before, expanding the catalog of genetic conditions and opening new treatment pathways.

Steps to Implement Near-Perfect Genome Sequencing in Clinical Practice

• Adopt long-read sequencing infrastructure: Clinical labs need to invest in long-read sequencing machines and train staff to operate and maintain them, shifting from the current short-read dominant approach.
• Build diverse pangenome databases: Ensure genetic reference data includes people from different ancestries and populations to reduce diagnostic bias and improve variant interpretation accuracy for all patients.
• Integrate AI-based variant interpretation tools: Deploy machine learning models trained to recognize novel disease-causing variants and prioritize which variants are most likely to explain a patient's symptoms.
• Establish bioinformatician and genomics collaboration: Foster partnerships between computational specialists and clinical genomicists to continuously improve and validate the near-perfect genome sequencing framework.

The research community is already moving in this direction. Hoischen and his team are calling on bioinformaticians and genomics researchers to adopt and improve near-perfect genome sequencing, recognizing that this is not a finished product but an evolving framework.

What Does This Mean for the Broader AI-Genomics Landscape?

The near-perfect genome sequencing framework represents a shift in how AI is being applied to biology. Rather than using AI only to analyze existing data, researchers are now using AI to interpret complex, high-resolution genomic information in real time. This mirrors a broader trend in biotech: generative AI models are being deployed not just to speed up research, but to improve the quality and accuracy of biological discovery.

At the BIO International Convention in 2026, industry leaders discussed how generative genomics, which uses AI to analyze genetic sequences for innovative treatments, could reshape drug development. The consensus was clear: AI's real value lies not in speed alone, but in improving success rates. Currently, 90% of clinical trials fail, a figure that has remained stubbornly stable for decades. Even a 5% improvement in success rates would be transformative.

However, experts emphasized that trust is essential. Julie Rytlewski, a Senior Director at Bristol Myers Squibb, stated: "If the most beautiful, accurate model doesn't change a decision, it's not useful. There has to be trust in what these models are doing, especially when it comes to treating people".

The near-perfect genome sequencing framework demonstrates this principle in action. By combining multiple data sources, using AI to interpret complex patterns, and maintaining rigorous experimental validation, researchers are building tools that clinicians and patients can trust. For rare disease patients who have endured diagnostic delays, this convergence of long-read sequencing, diverse genetic data, and AI-powered interpretation offers genuine hope for faster, more accurate answers.