New Tool Distinguishes Real Microbial Signals in Tumor Genomic Data

When scientists analyze tumor DNA, they often uncover snippets of bacterial, viral, and fungal genetic material. The lingering question is whether these microbes actually inhabit the tumors or are introduced during sample handling. Conflicting studies using the same datasets have produced opposing answers. Researchers at Rutgers Cancer Institute have addressed this uncertainty with a computational approach called PRISM—Precise Identification of Species of the Microbiome. Their work appears in Cancer Cell.

“There are microbes all over the environment, on our skin and in our breath,” said Subhajyoti De, senior author of the study. “There could be DNA particles floating in the air. How do you know what you're finding came from the tissue you were interested in, or was something introduced along the way?”

PRISM systematically reduces contamination risk by running a series of filtering and alignment steps. It removes residual human sequences, compares remaining genetic data to microbial reference databases, and uses a machine‑learning model to judge whether each detected microbe is authentic or an artifact. The tool can screen existing genomic data from humans to determine which microbial sequences likely represent real organisms in the original tissue.

Clarifying which microbes genuinely exist in tumors matters for studying treatment responses and for designing microbiome‑related therapy strategies. According to lead author Bassel Ghaddar, PRISM allows scientists to reuse standard sequencing data without extra laboratory expense.

To train and validate PRISM, the team assembled 833 samples from over 200 prior studies with established microbial content. The model achieved sensitivities and specificities above 90% and outperformed comparable analytical methods. Applied to roughly 4,400 tumors across 25 cancer types, PRISM revealed that microbial signals were strongest in cancers of the mouth, gut, and cervix—tissues normally exposed to microbes—while signals were minimal in internal organs, countering some earlier reports.

In pancreatic cancer, PRISM identified certain tumors containing E. coli, which can produce colibactin, a DNA‑damaging compound. These cases showed molecular patterns suggesting potential effects on tissue density and drug accessibility. “We are finding a signal in the bacteria that is correlated with a phenotype in the cells,” De said.