This research introduces PRISM, a novel computational framework designed to accurately identify and decontaminate microbial signals within human genomic data. Researchers developed this tool to resolve ongoing controversies in the cancer microbiome field caused by taxonomic misclassification and environmental contamination. By utilizing a gradient-boosted tree model and rigorous host-read removal, PRISM distinguishes truly present microorganisms from technical artifacts with high sensitivity and specificity. Analysis of TCGA and CPTAC datasets revealed robust microbial signatures in gastrointestinal, head-and-neck, and urogenital tumors, while finding only sparse signals in other cancer types. In pancreatic cancer specifically, the study linked the presence of microbes to greater smoking history and altered host protein glycosylation. Ultimately, PRISM enables the reliable re-evaluation of existing genomic archives to uncover meaningful host-microbe interactions in human disease.

References:

• Ghaddar B, Blaser M J, De S. Reliable detection of Host-Microbe Signatures in cancer using PRISM[J]. Cancer Cell, 2026.