Skip to main content
Premium Trial:

Request an Annual Quote

Microbiome-Based Cancer Dx Emerges From Cancer Genome Atlas Reanalysis

NEW YORK – Researchers led by the University of California, San Diego's Rob Knight have discovered unique microbial DNA signatures in the tissue and blood of cancer patients, leading to the development of a microbiome-based oncology diagnostic tool.

As the researchers described in a study published on Wednesday in Nature, they conducted a re-examination of whole-genome and whole-transcriptome sequencing data from 33 cancer types in The Cancer Genome Atlas (TCGA), analyzing sequence data from a total of 18,116 samples from 10,481 treatment-naive cancer patients, looking for microbial DNA reads.

The microbial signatures they found were not only unique within and between most major cancer types, but also remained predictive when applied to patients with stage Ia to IIc cancer, as well as cancers lacking any genomic alterations, measured on two commercial-grade cell-free tumor DNA platforms.

The researchers were also able to discriminate between samples from healthy, cancer-free individuals and samples from patients with multiple types of cancer solely using plasma-derived, cell-free microbial nucleic acids.

To advance this discovery toward regulatory approval, commercialization, and clinical application of a diagnostic test, Knight and first author Gregory Poore have filed patent applications, according to UCSD. They have also founded a spinout company called Micronoma, with study co-author Sandrine Miller-Montgomery.

"The ability, in a single tube of blood, to have a comprehensive profile of the tumor's DNA (nature) as well as the DNA of the patient's microbiota (nurture), so to speak, is an important step forward in better understanding host-environment interactions in cancer," study co-author Sandip Pravin Patel, a medical oncologist and co-leader of experimental therapeutics at Moores Cancer Center at UCSD Health, said in a statement. "With this approach, there is the potential to monitor these changes over time, not only as a diagnostic, but for long-term therapeutic monitoring."

The researchers looked at 4,831 WGS studies and 13,285 RNA-seq studies from the TCGA database. They used machine learning to identify microbial signatures that discriminated between types of cancer and then compared their performance. The trained machine learning models discriminated between and within types and stages of cancer, and the researchers found that the performance of these models was strong for discriminating one cancer type versus all others in 32 cancer types, as well as tumor versus normal in 15 cancer types.

In testing tissue-based microbial models for eight cancer types, the researchers found that the models performed well for discriminating between stage I and stage IV tumors for colon adenocarcinoma, stomach adenocarcinoma, and kidney renal clear cell carcinoma, but not the other five cancers, nor for discriminating intermediate stages. This suggested that microbial community structure dynamics may not correlate with cancer stages as defined by host tissue for all types of cancer.

The researchers also examined whether blood-based microbial DNA (mbDNA) could be clinically informative in cancer by applying their machine learning models to the WGS data from TCGA blood samples. They found that blood-borne mbDNA could discriminate between numerous types of cancer, regardless of the microbial taxonomic algorithm and database used for classification.

They further sought to benchmark their machine learning models against existing ctDNA assays, focusing on circumstances under which ctDNA assays fail: stage Ia–IIc cancers and tumors without detectable genomic alterations. After removing all blood-derived normal samples from patients harboring stage III or IV cancers, the researchers built new models and found that they were able to discriminate well between types of cancer using blood microbial DNA.

"We further used gene lists from the Guardant360 and FoundationOne Liquid assays to filter out TCGA patients with one or more targeted modifications and found that the same [machine learning] approach showed good discrimination for most remaining types of cancer," the authors wrote.

In a statement, Poore noted the study may prompt changes in the cancer biology research field.

"For example, it's common practice for microbiologists to use many contamination controls in their experiments, but these have historically been rarely used in cancer studies," he said. "We hope this study will encourage future cancer researchers to be 'microbially conscious.'"