Abstract 5713: Assessing the real-world utility of cell-free microbial DNA in diagnosing early-stage lung cancer

Abstract

Abstract Introduction: Links between cancer and microbes date back four millennia (Sepich-Poore et al. 2021. Science). Recently, we found that microbial DNA is detectable in tumor tissues and patient blood from many human cancer types (Poore et al. 2020. Nature). These intratumoral and bloodborne microbiomes were distinct between cancer types, between normal and malignant tissues, and present in cell-free plasma samples. However, the practical utility of cell-free microbial DNA (cf-mbDNA) as a bona fide liquid biopsy diagnostic, including its applicability in early-stage disease in treatment-naïve individuals, distinguishing histological subtypes, and discriminating against non-cancer-but-diseased patients remains unknown. Thus, we constructed an age and sex-matched cohort of >1000 individuals with lung cancer, lung disease, and no disease (healthy) to evaluate the utility of a cf-mbDNA-driven liquid biopsy diagnostic. Methods: Shallow shotgun metagenomic sequencing with gold-standard positive and negative controls was performed using 400 µL of patient plasma. Direct genome alignments separated human and microbial reads, and generated genome-wide binned and species-level abundances, respectively. Novel taxonomic diversity was captured by additionally performing de novo co-assemblies in tandem with tumor and blood samples from The Cancer Genome Atlas (TCGA). Multi-modal, stacked machine learning classifiers then evaluated the diagnostic performance of microbial-only and multi-species (microbial + human) information. Results: Cf-mbDNA provides strong diagnostic performance in treatment-naïve, cancer-bearing individuals versus age and sex-matched healthy controls, as early as stage I disease (AUROCs≥0.90). Furthermore, cf-mbDNA outperforms histological classification compared to human genomic information. Multi-species models paired with routinely-available clinicodemographic information provided robust discrimination of lung cancer versus lung diseases (AUROC≥0.80). Importantly, the addition of cell-free microbial information produced an integrated model surpassing the diagnostic performance of PET-CT and clinical risk models for lung nodule malignancy determination in a blinded validation cohort of Stage I lung cancer and non-cancer lung disease samples. Conclusion: Cf-mbDNA features comprise a novel class of biomarkers that are combinable with host analytes, and show promise for real-world, early-stage, lung cancer diagnosis. Citation Format: Serena Fraraccio, Stephen Wandro, Akanksha Singh-Taylor, Sandrine Miller-Montgomery, Eddie Adams, Rob Knight, Leopoldo N. Segal, Harvey I. Pass, Gregory D. Sepich-Poore. Assessing the real-world utility of cell-free microbial DNA in diagnosing early-stage lung cancer. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5713.