AI tool simplifies and scales complete genome assembly, supporting advances in diagnostics and precision medicine

An international research team led by the A*STAR Genome Institute of Singapore (A*STAR GIS) has developed HERRO, an artificial intelligence (AI) tool that could make it easier and more cost-effective to produce complete, high-quality genome assemblies.

HERRO corrects errors in nanopore sequencing reads generated by Oxford Nanopore Technologies (ONT) sequencers. ONT is a leading player in long-read sequencing, a technology increasingly used when scientists need to read very long stretches of DNA. These long reads help researchers study complex parts of the genome that shorter sequencing reads can find challenging, such as repetitive DNA, centromeres, and the difficult regions of the sex chromosomes.

In this study, the team focused on ONT Simplex reads, which are produced from a single DNA strand and have historically had higher error rates than some other sequencing technologies. HERRO addresses this limitation by using deep learning to correct errors in these reads, improving their accuracy by up to 100-fold. This allows researchers to generate high-quality genome assemblies using a single long-read sequencing platform, instead of relying on more complex workflows that combine multiple technologies. The paper is published in the journal Nature.

Why accuracy matters in complete genome assembly

A genome is the full set of genetic instructions in a living organism. For humans, it contains roughly six billion DNA “letters” across two sets of chromosomes, one inherited from each parent. Reading this blueprint accurately and completely helps scientists understand how genetic differences affect health, disease, evolution and biological diversity.

A key challenge in genome assembly is telling the difference between a sequencing error and a genuine biological difference. This is important because humans are diploid, meaning we carry two similar but non-identical copies of most chromosomes. If an error-correction tool is too aggressive, it may accidentally “smooth out” real differences between these inherited chromosome copies.

HERRO was designed to be haplotype-aware, meaning it improves the accuracy of ONT Simplex reads while preserving the genetic differences that distinguish one chromosome copy from the other. This helps researchers build more accurate and complete genome assemblies without losing information that may be biologically important.

What HERRO could unlock

By making nanopore reads more accurate and easier to use for complete genome assembly, HERRO could help researchers and industry users:

• Capture a more complete picture of the genome, including complex regions such as repetitive DNA, centromeres, and difficult regions of the sex chromosomes that can be challenging to study with standard approaches.
• Improve the detection of genetic variation, including structural variants relevant to inherited diseases, cancer and other conditions.
• Support precision medicine research, by improving genome maps used to study disease risk, diagnosis and treatment response.
• Scale genomics applications, by reducing the need for multiple sequencing platforms, separate sample preparations and large amounts of DNA, while supporting wider uses in agrigenomics, biodiversity research and biotechnology.

Building complete genome maps with a simpler workflow

When combined with state-of-the-art genome assembly methods, HERRO-corrected nanopore reads enabled the team to reconstruct complete, end-to-end human chromosomes, including the challenging X and Y chromosomes.

These are known as telomere-to-telomere, or T2T, genome assemblies. In simple terms, this means scientists can reconstruct a chromosome from one end to the other, with few or no gaps.

The team also showed that HERRO could deliver high-quality complete genome assemblies across multiple human genomes and several non-human organisms. These results were comparable to or better than assemblies produced using more complex multi-platform approaches, but achieved through a simpler workflow with lower DNA material requirements.

Senior author Prof Mile Šikić, Assistant Director (AI & Compute) at A*STAR GIS, and Faculty of Electrical Engineering and Computing, University of Zagreb, Croatia, said, “HERRO shows how AI can make ultra-long nanopore reads accurate enough for high-quality complete genome maps. This could make complete genome assembly simpler, more scalable and more cost-effective for research worldwide.”

First author Dr. Dominik Stanojević, Scientist at A*STAR GIS, and Faculty of Electrical Engineering and Computing, University of Zagreb, Croatia, added, “HERRO improves sequencing accuracy while preserving real differences between chromosome copies. This helps researchers build more accurate genome assemblies, especially in complex regions that standard methods may miss.”

Dr. Lakmal Jayasinghe, Chief Scientific Officer at Oxford Nanopore Technologies, said, “By leveraging HERRO-corrected nanopore reads, we demonstrate that high-quality, telomere-to-telomere genome assemblies can be achieved using a single sequencing technology, simplifying workflows while maintaining accuracy.”

Who’s behind this story?

Gaby Clark

MA in English, copy editor since 2021 with experience in higher education and health content. Dedicated to trustworthy science news.

Full profile →

Advertisements

Robert Egan

Bachelor’s in mathematical biology, Master’s in creative writing. Well-traveled with unique perspectives on science and language.

Full profile →