With the Human Genome Project being launched in the 1990s, researchers embarked on a mission to sequence the complete human genome. While they successfully completed the project in 2003, some crucial gaps remained.
Scientists turned to whole genome sequencing techniques which reduced these blind spots, but some gaps persisted. “In my teaching, sometimes, I put the ‘W’ of ‘whole genome’ in brackets,” said Alexander Hoischen, a genomic technologies researcher at Radboud University. “So, the idea [is] that for now, genome sequencing contains a lot of holes.”
This is especially relevant in diagnosing rare diseases wherein researchers must test for variants outside known disease-related genes. The absence of one method to read the complete human genome sequence leads to patients having to undergo a battery of expensive tests, but because of the remaining holes, these still do not guarantee a diagnosis.1
Alexander Hoischen, who researchers genomics technologies at Radboud University, employs long-read sequencing to diagnose rare diseases. Here, he is pictured next to some long-read sequencing machines.
© Radboud UMC
To fill this gap, over the past few years, Hoischen and his team explored how to supplement existing technologies to help researchers interpret novel rare disease variants. They developed a long-read sequencing test with higher genome coverage which helped conclusively diagnose more disease cases than conventional technologies did.2
Building on these results, the researchers proposed a framework called near-perfect genome sequencing, consolidating long-read sequencing with genome assembly from both chromosomes, more representative pangenome references, and AI-supported interpretation.3 This one-test paradigm could lead to faster and better rare disease diagnoses, potentially transforming the genetic diagnostics landscape.
Short-Read Sequencing Gaps Limit Rare Disease Variant Discovery
One of the most common methods in massively parallel sequencing is short read sequencing, wherein researchers decode DNA broken down into small fragments of about 100 base pairs.4 Bioinformatic tools help reassemble these pieces, but the short reads may not map to the entirety of the reference genome, leaving gaps and limiting the DNA sequencing technology’s application in identifying novel variants.
“Imagine the human genome being a jigsaw puzzle with a million pieces. The trouble with short reads is, [they are] tiny pieces,” explained Hoischen. Long-read sequencing, which involves reading fragments of up to 20,000 base pairs, could help overcome this challenge. “Long reads are just bigger pieces. So that's naturally much [more] intuitive and easy,” said Hoischen. He added that this method also examines epigenetic modifications which can turn genes on or off to cause diseases
To test the feasibility of long-read genome sequencing, Hoischen and his colleagues compared its outcome with that of conventional tests.3 Applying this sequencing to samples of more than 800 people with rare diseases yielded a conclusive diagnosis in 160 patients. In contrast, standard-of-care diagnostic testing—which involved multiple tests—helped diagnose only 137 patients, highlighting the utility of long-read sequencing as a single test for genetic diagnoses.
Integrating Pangenomics and AI to Achieve Rapid Genetic Answers
Despite this, Hoischen said that long-read sequencing along may not give the full picture. “The trouble with long reads only, with one technology, is…[the base pairs may not be] long enough to put the jigsaw puzzle completely together,” he said. An additional sequencing method could offer a fuller picture, but this shoots up the price of the diagnostic test.
A Single Molecule, Real-Time (SMRT) cell acts as the reaction chamber where millions of individual DNA molecules are sequenced simultaneously in real time.
© Radboud UMC
So, Hoischen and his team proposed using long-read sequencing as the foundation with diploid genome assembly as the representation of the individual genome. Coupling these technologies with pangenome data from diverse backgrounds could ensure equitable access to medical genomics, according to Hoischen.
While these methods together could help researchers identify novel rare disease-associated variants, interpreting it remains a challenge. AI-based variant interpretation tools could help scale the process.5 “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,” explained Hoischen.
Over time, Hoischen hopes that bioinformaticians and genomics researchers join forces to adopt and improve near-perfect genome sequencing, which could significantly improve rare disease diagnostics. “We would help our doctors so much more efficiently,” he said.
He added that methods like near-perfect genome sequencing could also bring unknown disease-causing variants into the spotlight, helping researchers identify more rare diseases. “We [could] help the patients that deserve clearer answers much more rapidly and much better. So that's a massive hope.”
- Seaby EG, Ennis S. Challenges in the diagnosis and discovery of rare genetic disorders using contemporary sequencing technologies. Brief Funct Genomics. 2020;19(4):243-258.
- de Bitter TJJ, et al. Clinical long-read genome sequencing for rare-disease diagnostics. N Engl J Med. 2026;395(4):405-408.
- Sabbagh Q, et al. Near-perfect genome sequencing in medical genetics. Nat Genet. 2026;58(7):1480-1489.
- Bleidorn C, et al. The untapped potential of short-read sequencing in biodiversity research. Trends Genet. 2026;42(2):137-149.
- Dias R, Torkamani A. Artificial intelligence in clinical and genomic diagnostics. Genome Med. 2019;11(1):70.
