Hello, my name is Eddie Adams, and it is a great pleasure to be able to present this poster discussing the use of circulating cell-free microbial nucleic acids for liquid biopsy-based detection of lung cancer. Microbes are the principal cause of numerous and anatomically diverse cancers, and metagenomic microbiome studies have provided ample evidence that our commensal microbiota influence cancer initiation, progression, metastasis, and anti-cancer drug metabolism. Given these associations, Micronoma's co-founders sought to determine the extent to which microbes may be directly associated with all cancers, irrespective of their potential role in disease causation, an exploration that led to the diagnostic modality for lung cancer we present here today. To accomplish this, Greg Poore and colleagues analyzed all treatment-naive whole-genome and transcriptome studies from the Cancer Genome Atlas for their bacterial, viral, and archaeal nucleic acid content. TCGA dataset consisted of over 18,000 whole-genome and RNA sequencing samples from 10,481 patients across 33 different tumor types, and included non-neoplasmic tumor-adjacent tissue in patient blood samples. Aligning the TCGA sequencing reads to a reference human genome revealed that a significant fraction, 7.2%, were of non-human origin, of which 35.2% could be taxonomically assigned to bacteria, viruses, or archaea. Here you see the percentage of microbial reads across TCGA cancer types, with lung squamous and adenocarcinoma highlighted. The taxonomically assigned datasets were used to train machine learning models using a 70-30 trained test split to discriminate between and within types and stages of cancer. The trained models could effectively discriminate one cancer versus all others, and tumor versus normal adjacent tissue, solely using microbial DNA or microbial RNA. Here we show some examples of microbial DNA-based cancer classification. In one, ovarian tumor microbes differentiated ovarian cancer from all other TCGA cancers. In two, microbial signatures from blood-derived normal samples differentiate lung adenocarcinoma from all other cancers with matched blood samples. In three, blood-borne microbial DNA signatures correctly classify low-grade colon adenocarcinoma. And finally, in four, differentiation of primary breast cancer tumors from normal adjacent tissue was done with high-performance classification. To further validate this finding and extend it to a more standardized liquid biopsy format, plasmid-derived cell-free microbial DNA from 25 patients with stage 3 and 4 lung cancer were compared to cell-free microbial DNA from 69 HIV-negative healthy patients. Using the same microbial detection pipeline and machine learning steps as before, cell-free microbial DNA-based signatures resulted in accurate discrimination between lung cancer versus healthy samples, with 22 out of 25 lung cancer samples correctly classified with five false positives, yielding a sensitivity and specificity of 88 and 93 percent, respectively. Data which demonstrates that microbial nucleic acids are powerful biomarkers, wholly orthogonal to the human genome, the cancer genome, that can be hardest for liquid biopsy disease detection. The Micronoma's mission is to employ cancer-associated cell-free microbial DNA signatures for diagnostic applications with a first focus on lung cancer, for which effective early-stage diagnostics are desperately needed. It would be critical to expand the breadth of controls to include non-healthy, non-cancer complications of the lung, such as chronic obstructive pulmonary disease, sarcoidosis, interstitial fibrosis, and other conditions where microbial involvement may overlap substantially with that of lung cancer. This is something that we are working on presently. Thank you very much for your interest.

circulating cell-free microbial nucleic acids

liquid biopsy

lung cancer

microbial DNA

machine learning models