Interview: Detecting, classifying, and monitoring CNS tumors with nanopore sequencing
Carlo Vermeulen is a postdoctoral researcher at UMC Utrecht Center for Molecular Medicine and Oncode Institute, where he focusses on the use of nanopore sequencing to detect DNA methylation in cancer. Ann-Kristin Afflerbach is pursuing her doctorate at the Research Institute Children’s Cancer Center Hamburg Research, where she works on utilising liquid biopsies for use in diagnostics in pediatric oncology.
We caught up with Ann-Kristin and Carlo to talk about how they became interested in molecular genomics, their latest research using nanopore sequencing, and the challenges they face. Carlo also shared further insight on providing rapid, molecular classification of CNS tumours during surgery, whilst Ann-Kristin discussed characterising tumours from low input samples.
You can hear more about Carlo's and Ann-Kristin’s work in the recent webinar ‘Detecting, Classifying, and Monitoring CNS Tumors With Nanopore Sequencing'.
What are your current research interests?
Carlo: The lab I work in focusses mostly on artificial intelligence (AI), with me as an oddball molecular biologist with some interest in bioinformatics. We use nanopore sequencing for many projects, and I really like all the exciting new possibilities it offers, such as direct detection of DNA modifications. My main focus has been to leverage our expertise in AI and apply these new technological advances in cancer research.
Anna-Kristin: I am currently focusing on diagnosing and monitoring brain tumors using liquid biopsies, with the aim to have a molecular diagnosis of the tumor without the need for surgical biopsy. This will be especially important in cases with difficult-to-reach anatomical sites.
What first ignited your interest in molecular genomics and what lead you to focus on oncology?
Carlo: I started my PhD working on fundamental biology, and whilst I found the subject interesting, I also really liked the technological and computational aspects of it. I ended up spending a large part of my PhD applying technology from fundamental biology for diagnostic applications which came with a lot of new and particular challenges. I found I really enjoyed the combined effort between bioinformatics, wet lab technology, and clinicians.
When I started as a postdoc, nanopore sequencing was an up-and-coming technology, but it was immediately obvious that it would enable a lot of exciting new possibilities for clinical applications, especially in cancer research where DNA methylation is a key aspect.
Anna-Kristin: I was initially interested in the brain and came to oncology by chance, but now I find it quite fascinating that we can distinguish so many tumor entities by molecular characteristics and methylation signatures.
How has nanopore sequencing benefited your research into the role that methylation plays in cancer?
Carlo: Our work using nanopore sequencing for CNS tumor classification during surgery is a nice example of technology-driven research. We were familiar with nanopore sequencing and methylation detection from earlier projects and when we met with the clinicians that treat CNS cancer patients, we learned of their experience with methylation-based profiling. It was immediately clear that intraoperative nanopore sequencing could potentially make a big impact on surgical outcomes.
Anna-Kristin: It has allowed us to perform methylation analysis as well as sequence analysis from the same sample, without having to choose between methods. This is especially valuable when sample amounts are little and precious.
What impact could the ability to accurately determine copy number changes and methylation profiles from liquid biopsy samples potentially have for the treatment of brain cancer?
Carlo: Brain cancers are obviously a very challenging type of cancer to treat as the tumors are difficult to access and are also surrounded by brain tissue. For resection surgery, there is a very difficult choice between radically resecting the tumor with a risk of severe side effects, versus conservatively resecting the tumor to reduce the risk of side effects but leaving some of the tumor in place. To make this choice, the surgeon would ideally know exactly what kind of tumor they are dealing with, but at present, the final molecular diagnosis only comes a week or so after surgery. This sometimes reveals that another surgery is needed, or that the surgery could have been more conservative.
We hope that in the future we can perform intraoperative sequencing alongside our neural network classifier ‘Sturgeon’ to provide a full molecular diagnosis from DNA methylation profiles during surgery, therefore providing time to tailor the surgical plan to the tumor type.
Anna-Kristin: It will help to provide the best of care from the moment brain cancer is suspected, right through to influencing surgical strategies and early detection of relapses.
What have been the main challenges in your work and how have you approached them?
Carlo: One of the biggest challenges is training an AI to classify many different subclasses in a relatively rare disease. Ideally, we would like to train and test our AI on hundreds of cases from every different class of tumor, but it would take decades to collect enough samples.
We addressed this by simulating shallow nanopore sequencing experiments from deep sequencing experiments or methylation arrays. That way we can train and test our classification AI on millions of simulated samples.
Anna-Kristin: Some of the main challenges have been the small amount of DNA that we were able to isolate, as well as nanopore sequencing being a method initially designed for long reads and so only produces small amounts of data. We have had to perform additional size selection steps within the wet lab steps and optimize our bioinformatics workflow to work with smaller fragments, and reduced amount of data produced from our experiments.
What’s next for your research?
Carlo: Hopefully we can inspire a lot of clinics to get this tool set up at their own clinics and so learn more about the performance and challenges in different settings. The next step for us now is clinical validation and deciding how to incorporate our Sturgeon classification workflow into clinical practice to potentially make intraoperative sequencing part of the clinical routine.
Anna-Kristin: Next is to potentially integrate our workflow as an additional option for routine diagnostic labs, to enlarge our cohort, and eventually expand to other entities as well!