Scientists at the A*STAR Genome Institute of Singapore have unveiled HERRO, an artificial intelligence tool that corrects errors in DNA sequencing with stunning precision—improving accuracy by up to 100-fold and simplifying the process of reading an entire human genome from end to end.
The breakthrough matters because understanding our genetic blueprint completely and accurately underpins everything from diagnosing inherited diseases to developing personalized medicines. A human genome contains roughly six billion DNA "letters" across two sets of chromosomes, and the challenge has long been distinguishing between a genuine biological difference and a mere reading error. HERRO, trained using deep learning, solves this by correcting mistakes in nanopore sequencing reads—long stretches of DNA that are produced by Oxford Nanopore Technologies sequencers, a leading platform for reading particularly tricky genomic regions.
These long reads are especially valuable for studying complex parts of the genome that shorter sequencing technologies struggle with: repetitive DNA, centromeres, and the difficult regions of the sex chromosomes. The team focused specifically on ONT Simplex reads, which are generated from a single DNA strand and have historically carried higher error rates than competing approaches. By teaching HERRO to recognize and fix these errors while preserving real biological differences between the two chromosome copies humans inherit, researchers created a tool that works like a highly skilled editor—catching typos without changing the intended meaning.
What makes HERRO haplotype-aware—the technical term for this preservation of meaningful genetic differences—is crucial. Humans are diploid, meaning we carry two similar but distinct copies of most chromosomes. An overly aggressive error-correction tool might smooth out these real differences, losing information that could be biologically important for understanding disease or evolution. HERRO was designed to avoid this pitfall, maintaining genetic nuance while dramatically improving read quality.
The practical payoff is substantial. When combined with state-of-the-art assembly methods, HERRO-corrected nanopore reads enabled the team to reconstruct complete human chromosomes from end to end—what scientists call telomere-to-telomere, or T2T, assemblies. This means chromosomes reconstructed with few or no gaps, including the notoriously difficult X and Y sex chromosomes. The team demonstrated that HERRO delivers high-quality complete genome assemblies not just for humans but across multiple organisms, achieving results comparable to or better than those from far more complex, multi-platform approaches—yet requiring simpler workflows, less DNA material, and fewer preparation steps.
The implications ripple outward across medicine and science. More complete genome maps could sharpen detection of structural variants linked to inherited diseases and cancer. In precision medicine, where treatment decisions increasingly depend on individual genetic profiles, HERRO could accelerate the development of tools that predict disease risk and treatment response. Beyond human health, the tool opens doors for agrigenomics, biodiversity research, and biotechnology—fields where access to high-quality complete genomes has been limited by cost and complexity.
By making nanopore sequencing more accurate and easier to use, HERRO transforms what was once a specialized workflow into something more accessible and scalable. The work, published in Nature, suggests that the future of genomic science may be simpler and more elegant than the present.