Researchers from FER and the Genome Institute of Singapore (GIS, A*STAR), in collaboration with Oxford Nanopore Technologies, have developed HERRO, an artificial intelligence-based tool that enables complete reconstruction of the human genome using only a single long-read DNA sequencing technology.

The results were published in Nature, one of the world’s most influential scientific journals, which features outstanding discoveries in biology, physics, medicine, chemistry, and other scientific fields.

The co-authors of this exceptional achievement are Prof. Mile Šikić (GIS, A*STAR Singapore and FER), Dominik Stanojević, postdoctoral researcher (GIS, ASTAR Singapore and FER), Dehui Lin, PhD candidate (NTU Singapore), Sergey Nurk, Chief Bioinformatician (Oxford Nanopore Technologies), and Paola Florez de Sessions, Deputy Director (Oxford Nanopore Technologies).

The HERRO tool significantly reduces error rates in DNA fragment readings while minimizing the loss of real genetic differences between the two copies of chromosomes inherited from parents. To achieve this, HERRO reduces sequencing errors and distinguishes true genetic variation between the two chromosome copies from measurement errors. HERRO compares multiple overlapping reads of the same genomic region, the complete genetic material of an organism, and uses them to predict the correct base at each position. The result is up to a hundredfold increase in read accuracy.

"We have shown that reads corrected by our method enable reconstruction of human chromosomes, including the challenging X and Y chromosomes. In typical human genome cases, more than 30 out of 46 chromosomes are reconstructed without any gaps", explains Prof. Šikić from FER.

The human genome contains around six billion base pairs, letters of the genetic code, organized into 23 pairs of chromosomes, with one set inherited from each parent. A complete and accurate reconstruction of this code is crucial for understanding inherited diseases, developing new drugs, and advancing personalized medicine.

This innovation by Croatian scientists opens the path toward full genome reconstruction at lower cost and with a simpler laboratory workflow.
 

Biographies of Croatian co-authors

Dominik Stanojević, PhD, completed his studies in computer science at the Faculty of Electrical Engineering and Computing, University of Zagreb, where he also earned his PhD. During his doctoral research, he completed a research stay at the Genome Institute of Singapore (GIS, A*STAR), where he continues to work as a scientist. His research focuses on applying artificial intelligence in genomics, particularly genome reconstruction and detection of epigenetic modifications. He is the author of four journal papers and one conference paper.

Prof. Mile Šikić, PhD, is a researcher in computational genomics and artificial intelligence. He is currently a full professor at the Faculty of Electrical Engineering and Computing (FER), University of Zagreb, and a group leader at the Genome Institute of Singapore (GIS), A*STAR.

He earned his PhD in computer science from the University of Zagreb in 2008 and has since been recognized for his work in developing algorithms and applying AI in genomics. His work includes de novo genome assembly (Racon, Raven), error correction in nanopore sequencing (HERRO), and large language models for RNA (RiNALMo).

Prof. Šikić leads a research team of more than 15 scientists in Singapore and Croatia, promoting interdisciplinary collaboration. He is an active member of several international consortia, including T2T, HPRC, and Cancer Genome in the Bottle.

Early in his career, he worked as a systems integrator, consultant, and project manager on more than 70 industrial projects in computer networks, mobile networks, and cybersecurity. He also worked on complex and social network analysis, co-developing a new methodology for predicting electoral and market trends, which successfully predicted Donald Trump’s victory and Brexit in 2016. Using the same methodology in 2023, he co-founded Oraclum Capital, a hedge fund based in New York, with two partners.