By Benjamin Ross
March 24, 2020 | What if there was a way to determine who has cancer by simply analyzing microbial DNA patterns in blood samples?
This was a question Gregory Poore, an MD PhD student at the University of California, San Diego (UC San Diego) School of Medicine, posed to Rob Knight, professor and faculty director of the university’s Center for Microbiome Innovation, in 2017. After preliminary dive in data analysis, Knight put Poore in 2018 in contact with Sandrine Miller-Montgomery, executive director at the Center for Microbiome Innovation and Professor of Practice in Bioengineering at UC San Diego, who tells Diagnostics World News that she was immediately struck by the potential of Poore’s question and the method that was subsequently developed mid-2019.
“[Poore] wanted to see if what he had done was worth patenting,” Miller-Montgomery, who spent almost two decades in industry R&D and then heading sales, marketing, and product management at various biotech companies prior to joining UC San Diego, recalled. “The least I could say was, ‘Oh my gosh, this is amazing!’ It was obviously too good to be true, so I said, ‘Let’s try to break it.’ But once we tried several times to break it with no success, I suggested we fast track filing for Intellectual Property (IP).”
Fast forward to 2020 and, with Miller-Montgomery’s assistance, Poore’s method to identify cancers in microbial DNA has been spun out into Micronoma, a startup working toward clinical application , regulatory approval, and commercialization of a diagnostic test. The method has also been detailed in a recent Nature study (DOI: https://doi.org/10.1038/s41586-020-2095-1). Poore, Miller-Montgomery, Knight, and the other study authors hope this new diagnostic approach can change how cancer is both viewed and diagnosed.
“It’s opening the door to a couple of avenues,” said Miller-Montgomery, though there are still questions when it comes to the biological reasons for presence of microbes within cancers. “We don’t know yet what these microbes are doing in cancer tissue. We don’t know yet if they are recruited by the tumor in order to create an ecosystem for the tumor to grow, or if they are sent by the immune system as a way to defend itself against the tumor, or maybe they’re doing something else entirely. But we’re opening the door to say, Look over here. There are many discoveries to be made here.”
Poore and the rest of the study’s authors first re-examined whole-genome and whole-transcriptome sequencing studies in The Cancer Genome Atlas (TCGA). Focusing on 33 cancer types, such as colon cancer and prostate cancer, they analyzed over 18,000 samples, asking themselves if it was possible to find these unique microbial signatures in tissue and blood among the cancer types.
“The answer was yes, we could detect different microbial signature data based on the type of cancer in tissue,” Miller-Montgomery said. “At the same time, what was awesome to us was that all of the samples we were analyzing were treatment-naïve, meaning none of the samples we were looking at had received treatment. So we knew that treatment wasn’t explaining what we were seeing.”
The authors turned to machine learning for the initial analysis, developing a model that would associate microbial patterns with the presence of specific cancers in tissue and even in blood.
“The number of data that are in the database we were looking at took six months on a massive cloud server, computing full-time, to be able to analyze the full dataset,” Miller-Montgomery said. “It would be impossible to do this by the human eye on a regular spreadsheet editor.”
Following their analysis of samples from TCGA, Miller-Montgomery says the team shifted their focus to analyzing new samples from, from 100 patients at Moores Cancer Center at UC San Diego Health via a minimally invasive blood draw. They also collected samples from healthy volunteers for comparison.
What the team found was that their models were able to distinguish people with cancer with 86% accuracy from those without cancer.
That’s not to say the method is fool proof. Miller-Montgomery and her fellow researchers state that there’s still the possibility of false-positives or missed cancers from blood-based microbial DNA analysis. Further refinement of the machine learning models will reduce limitations, they say.
The blood draw is in stark contrast to typical cancer diagnostic procedures, including invasive and time-consuming biopsies. Miller-Montgomery says Micronoma’s solution is a repurposing of existing technologies, where analysis would be conducted on a sequencing machine, similar to what you would use on another assay such as the traditional liquid biopsies who are typically focusing on human and tumor DNA markers. But this time the markers are microbial ones.
“The good thing that plays in our favor is that, in comparison to a very low number of mutations, we are looking at a multitude of signatures,” she said. “So it’s not as if we are looking for one bacteria or one particular virus floating around in the blood; it’s a combination that gives us the answer, which increases the likelihood of us being able to do early detection.”
Miller-Montgomery says the business model of Micronoma is similar to its technology. “We’re not fully trying to reinvent the wheel,” she said. “We are just optimizing it and putting it on track toward markers that may be way more insightful.”
Micronoma’s assay will roll out in two phases, starting with a series of laboratory developed tests in which samples will be run at Micronoma CLIA lab, who will then send back an analysis report to the prescribing clinician, and in the future adding standard kits to its portfolio so that can be sent to labs all over the world.
While still a few years from making diagnostic kits a reality, Miller-Montgomery says they are working full speed ahead to discover new ways to diagnose cancers quicker and easier.