AI Improves Cancer Mutation Detection

A research team led by Professor Ruibang LUO, Assistant Director (Learning Experience & Student Enrichment) and Associate Head of the Department of AI & Data Science, School of Computing and Data Science (CDS), The University of Hong Kong (HKU), has developed ClairS—a novel deep-learning algorithm that significantly improves the detection of cancer mutations using long-read sequencing. Tested on breast cancer, lung cancer and melanoma cell-line datasets, ClairS has demonstrated exceptional accuracy across various cancer types and sequencing conditions.

Cancer mutations are genetic changes found in tumour cells but not in healthy cells; detecting them accurately is essential for cancer research and precision medicine. However, many existing methods are designed for short-read sequencing and often struggle to analyse structurally complex regions of the human genome. ClairS addresses this challenge by utilising long-read sequencing to reveal mutations that might otherwise be missed.

A key innovation of ClairS lies in its method for obtaining training data and labels. Because high-quality cancer training data is limited, the research team designed a novel strategy that mixes sequencing data from normal human samples to generate synthetic tumour-normal data. This allows ClairS to generate a virtually unlimited number of realistic, cancer-like training examples covering various tumour purities, sequencing depths and mutation levels. The result is a flexible and robust artificial intelligence (AI) model for real-world cancer genomic analysis.

ClairS has been tested on multiple cancer cell lines—including breast, lung, melanoma and pancreatic cancers—demonstrating high accuracy in detecting small cancer mutations. Importantly, ClairS has already been integrated into Oxford Nanopore Technologies’ official somatic variant-calling workflow, making it part of a practical commercial analysis pipeline. This marks a major step towards the widespread application of advanced sequencing technologies in clinical genomics.

Professor Luo commented, "Long-read sequencing is transforming how we study cancer genomes, especially in regions that were previously difficult to analyse. ClairS makes it possible to train powerful AI models even when real cancer training data is limited, supporting more reliable cancer mutation discovery from long-read sequencing data."

This study represents a significant step towards more accurate and comprehensive cancer genomic analysis. It also highlights a scalable approach to training medical AI when real clinical data is scarce, laying a solid foundation for the future of long-read clinical genomics.

Reference: Zheng Z, Chen L, Su J, et al. ClairS: a deep-learning method for long-read tumor–normal pair somatic small variant calling. Nat Methods. 2026:1-14. doi: 10.1038/s41592-026-03152-4

This article has been republished from the following materials. Note: material may have been edited for length and content. For further information, please contact the cited source. Our press release publishing policy can be accessed here.