A team of researchers at the University Medical Center Schleswig-Holstein (UKSH), Kiel Campus, the Kiel University, and the Max Planck Institute for Molecular Genetics, Berlin, have developed an innovative method for real-time molecular genetic classification of brain tumors during surgery. This approach combines DNA methylation analysis with advanced machine learning technologies to provide detailed information about the tumor type during surgery.

For the first time, neurosurgery can be tailored not only to the location and proximity of a tumor to critical brain functions but also to its specific molecular characteristics. This enables a targeted approach to tumor removal, which is a decisive advantage for patients, especially in complex cases. The findings were published on 28 February 2025 in Nature Medicine.

"Only through close collaboration between basic research scientists and physicians working in translational medicine was it possible to develop a method that surpasses all previous similar approaches in terms of precision and speed," says Prof. Dr. Franz-Josef Müller, deputy Director of the Department of Psychiatry and Psychotherapy at the UKSH, Kiel Campus, and Professor in the medical faculty of the CAU.

He led the interdisciplinary team together with Prof. Dr. Helene Kretzmer, former group leader "Computational Genomics" at the MPIMG, Dr. Alena van Bömmel, former postdoc at the MPIMG, Ph.D. student Mara Steiger, MPIMG, and Ph.D. student Björn Brändl, Center for Integrative Psychiatry, Campus Kiel.

The study was conducted in collaboration with the Department of Neurosurgery at UKSH, Campus Kiel. "While past advances in neurosurgery have refined techniques for tumor removal, this disease-centered approach revolutionises the entire procedure. Instead of simply focusing on the tumor's location, intraoperative identification of the tumor type opens up new possibilities for real-time adjustments in the surgical strategy based on the tumor's specific biology.

"This approach directly links precision medicine with neurosurgery, elevating treatment from a technique-driven procedure to a personalised, disease-centered therapy. Surgery, in this way, becomes a custom-tailored therapy," says Dr. Carolin Kubelt-Kwamin, senior physician at the Clinic for Neurosurgery.

DNA methylation, an epigenetic marker, serves as a "fingerprint" that reflects a tumor's origin. Different types of tumors exhibit distinct methylation patterns, which can be analysed via sequencing. The new method employs nanopore sequencing to identify these patterns with remarkable speed and accuracy.

"However, because this sequencing must occur within the narrow timeframe of surgery, it can realistically only capture a fraction of the tumor's methylation patterns," explains Dr. Kretzmer. To address this limitation, the researchers applied a mathematical approach from probability theory known as Bayes' theorem.

"By using this method, we trained a machine learning model that, in combination with specialised software, processes the sequencing data in real time and enables tumor classification in less than an hour using less than 0.1% of the genetic data," says Mara Steiger. This represents a significant advancement over previous methods, which took hours to weeks.

The study demonstrates that the results from this new method align with those of a complete neuropathological examination. The researchers highlight that the method can accurately classify even diagnostically challenging tumors, where traditional histopathological techniques often fall short. This innovation offers a significant advantage for patients, particularly in complex cases, by enabling a more targeted and precise approach to tumor removal.

Modern molecular and epigenetic analyses have revealed that many tumors previously thought to be one type of cancer are, in fact, highly diverse. For example, tumors of the central nervous system are now classified into nearly 90 distinct categories. These differences demand different therapeutic strategies. While some tumors can be treated with radiation or medication, others require extensive surgery.

Until now, however, tumor tissue was only analysed postoperatively, leaving surgeons uncertain about the tumor's classification during surgery and forcing them to make decisions that risk damaging healthy brain tissue. Intraoperative DNA methylation analysis, combined with nanopore sequencing, now provides critical information that surgeons can use in real time to make more informed decisions, supporting personalised precision surgery.

More information: Björn Brändl et al, Rapid brain tumor classification from sparse epigenomic data, Nature Medicine (2025). DOI: 10.1038/s41591-024-03435-3