Gregory Poore, a PhD student at the University of California San Diego School of Medicine, became interested in the study of microbes and cancer after his grandmother was diagnosed with late-stage pancreatic cancer, despite showing no symptoms. In 2017, Poore saw a study published in Science that showed how microbes invaded the majority of pancreatic cancers and break down the main chemotherapy drug that is given to these patients.

Poore is currently conducting his thesis in Rob Knight’s lab, professor and Director of the Center for Microbiome Innovation, where together with an interdisciplinary group of collaborators, they have developed a novel method to identifying who has cancer, and which type, by analysing microbial DNA in blood samples.

The study, published in Nature yesterday, may change how cancer is diagnosed, and shape our understanding of how cancer interacts with microbes.  They first looked at microbial data from The Cancer Genome Atlas and the team found associations that are already well documented, e.g. human papillomavirus (HPVO) and cervical cancer, but they also identified previously unknown microbial signatures that strongly discriminated between cancer types.

The researchers then trained and tested machine learning models to associate the patterns with the presence of specific cancers to be able to identify a patient’s cancer type from the microbial data in their blood. They also removed high-grade cancers and found that these cancer types were still distinguishable in the early stages of cancer, indicating this could be used for early detection.

The team tested the model by analysing blood samples from different cancer types and found that it was able to distinguish most people with cancer from those without.

Currently, most cancers require a tissue sample via invasive procedures. Liquid biopsies have been an area of interest for a long time, but currently, liquid biopsies are unable to distinguish normal genetic variation from early cancer and cannot pick up cancers where human genomic alterations aren’t known or not detectable, giving the risk for false-negative results.

Although an interesting discovery on the way microbes interact with cancers, the researchers point out there is still a possibility that blood-based microbial DNA readouts could miss cancer and give a false-negative result. However, they expect their new approach will become more accurate when they feed more data into their model.