Annually in the Netherlands, around 1,400 adults and 150 children are diagnosed with a tumor in the brain or spinal cord. While surgery is often the first step in treatment, neurosurgeons do not know precisely what type of brain tumor they will encounter and what degree of aggressiveness they are dealing with. Ideally, the exact tumor type needs to be determined before resection. Currently, however, the required diagnosis is only available after the tumor tissue has been visually and molecularly analyzed by a pathologist – a process that can take up to two weeks.

Oncode Researchers from the de Ridder laboratory at UMC Utrecht in collaboration with researchers from the Princess Máxima Center and Amsterdam UMC have developed a new algorithm that significantly speeds up the process. They recently published the results of their research in Nature. The algorithm, called ‘Sturgeon’, is based on Artificial Intelligence (AI). Using it, clinicians can now analyze cancer tissue during surgery and determine the tumor type in under two hours, allowing neurosurgeons to adjust their surgical strategy accordingly.

“As a bioinformatician, I’m trained to do interdisciplinary research and act as a connecting person, constantly speaking two languages. We had been working for some time with a new technology called Nanopore sequencing to speed up and improve analysis from liquids such as blood. With this technology, it is also possible to detect methylation patterns in DNA. The analysis is much faster than the current microarray platform used for routine diagnoses – it takes half an hour, compared to a week.

Then I got a phone call from Bas Tops, head of diagnostics at the Princess Máxima Center for Pediatric Oncology. He told me he could see the benefit of Nanopore sequencing for children with brain tumors.”

“There are more than 80 different possible brain tumor types. The classification dictates the amount of tissue the surgeon should remove – ideally as little as possible to prevent side effects such as memory loss or speech impairment. Using microarrays, we had to wait one to two weeks for a classification. That either meant the patient would have to go into surgery a second time, or the neurosurgeon would take an aggressive approach to removing brain tissue, to make sure the tumor wouldn’t return.

Bas and I started thinking about the premise of a quicker analysis using Nanopore sequencing. If we could diagnose a patient during surgery, the surgeon could act on tumor characteristics directly and remove brain tissue accordingly. We felt we were onto something.”

“To distinguish these tumor types based on DNA methylation patterns, we needed to develop a so-called classifier – an algorithm. During surgery, we have about an hour to sequence the tumor and determine the DNA methylation profile. That leads to a very incomplete picture. Because of the sparsity of such data, we couldn’t train our classifier well enough. That’s why we decided to develop a computer program simulating Nanopore data, based on the abundantly available microarray data. The sparse data turned out to be a blessing in disguise. There isn’t much overlap, allowing us to simulate multiple Nanopore ‘samples’ based on only one microarray sample. In the end, we were able to simulate 45 million samples based on earlier data.

Our lab had the Nanopore up and running while the Princess Máxima Center took care of the samples. Because of Oncode Institute’s base funding, we could quickly act on this opportunity. A PhD student and a postdoc in my group were willing to invest the time and generate the data for a proof-of-concept.”

“First, we had to validate it on a bigger scale. That process takes more resources. Oncode Institute’s technology development fund turned out to be an ideal instrument, allowing us to be up and running within only a few months. With the resulting data, we successfully applied for a large children’s cancer (KiKA) grant, delivering the means to validate our method even more thoroughly. That led to a timely submission to Nature. Looking back, I think it was a quick and painless journey.

Oncode Institute accelerated the process leading to publication of our results. It’s so different from the traditional way – respond to a call, write a proposal, wait for the money to hire someone, and reach all milestones before you can eventually publish your results. Thanks to Oncode Institute’s base funding we could take our first steps unhampered, followed by an extra nudge in the right direction with the technology development fund. Oncode Institute recognized the fact that we had struck gold and supported us along the way. Thanks to that speed, patients in the Princess Máxima Center can profit from our research already.”

“We’re trying to set up a prospective trial proving the difference between standard care and a new protocol using our technology. We’re still in the pilot phase, so we need to properly validate it. Luckily, we can count on the cooperation of the surgeons working in the Princess Máxima Center. Their willingness has led to a protocol that they’re already using. It affects brain surgery in children every week.

The nice thing about our concept is that we have proven we’re able to translate old data into new data. It simply can’t be true that historical patient data will become useless whenever we decide to switch measurement technologies. Currently, we have a patented method that we’re expanding to other tumor types. If we can rapidly characterize sarcomas and leukemias too, it’s possible we can reduce the time to diagnosis and the stress that comes with waiting for a treatment. Our research has a positive effect on patient care, that’s for sure.”