Deciding how aggressively to remove a brain tumor depends a lot on the tumor type. For some types, the decreased chance of further cancer development is worth the higher risk of brain damage from a more aggressive surgery. However, tumor type is hard to identify beforehand. If doctors belatedly determine that a patient’s tumor needed more aggressive removal, the patient will have to go in for a second surgery. A new procedure avoids this issue by using DNA sequencing and machine learning to quickly classify brain tumors while the surgery is still happening.

To classify the tumors, the research team looked at a chemical property of the tumors’ DNA called the methylation profile. Different types of tumors have different methylation profiles, and new kinds of DNA sequencing machines can identify this profile much faster than older methods. The research team gathered old data from tumor samples and created a machine learning algorithm to identify tumor type based on methylation profile. They then put their process into practice for 25 patients in surgery; after doctors removed each tumor, they read its methylation profile via DNA sequencing and classified the tumor type, all in less than 90 minutes. Out of 25 tumors, the procedure made 18 classifications, all of which were correct. For a few patients, the tumor information would have been valuable for informing the course of the surgery.

No algorithm is perfect. The authors point out that their algorithm is based on data from older technology, which produces less comprehensive methylation profiles. They hope this technique will improve over time as technology progresses. The quality of the results also hinge on the purity of the tumor sample that the surgeon extracts. Nonetheless, this procedure is an exciting confluence of rapidly developing technologies that benefits patients.

The research team includes Carlo Vermeulen and Marc Pagès-Gallego, a postdoctoral researcher and PhD student, respectively, in the lab of Jeroen de Ridder, an associate professor at UMC Utrecht in the Center for Molecular Medicine, and Bastiaan Tops, the head of the pathology lab at the Princess Máxima Center for Pediatric Oncology.

Managing correspondent: Alex Yenkin

Original Journal Article: “Ultra-fast deep-learned CNS tumour classification during surgery,” Nature

Press Article: “Rapid Nanopore Sequencing, Machine Learning Enable Tumor Classification During Surgery,” GenomeWeb

Image Credit: Rawpixel

Leave a Reply

Your email address will not be published. Required fields are marked *