Abstract Background Shotgun sequencing-based metagenomics is a useful approach to profiling microbiomes in environmental and patient samples. In a clinical setting, metagenomic techniques have the advantage of identifying organisms, which cannot be readily cultured or confirmed by other techniques. We have developed a clinical-grade, streamlined metagenomics-based pipeline that includes regulatory compliant method considerations, such as an internal control followed by a machine-based learning (ML) process to identify pathogens in urine samples. Methods We built an optimized novel end-to-end NGS assay pipeline that harnesses pathogen-specific genome data to detect bacterial species. We processed de-identified clinical urine specimens, collected from patients symptomatic for urinary tract infection (UTI). This workflow includes an IPC, QIACube-MDx extraction, library preparation and Illumina NextSeq 550 sequencing and a novel interpretable ML based analytic approach, Biotia-DX. Clinical culture results and qPCR were used as a baseline for the assay to train the ML model and to establish accuracy relative to the clinical standard of care. Results We clinically validated over 40 key uropathogens and conducted clinical studies of specificity, intra/inter reproducibility, accuracy in urine specimens (n=300), and limit of detection in E. coli, K. pneumoniae, P. mirabilis, S. aureus, E. faecalis and Candida. Additionally, the implementation of an internal control coupled with our Biotia-DX software provides an accurate (F1 score 94.3%) and highly sensitive clinical grade diagnostic tool. Conclusion Urine has historically presented a challenge for diagnostics via culturing, with a high rate of culture-negative results (∼30% on average). We improved the clinical utility of an NGS urine assay by leveraging an IPC and ML software. This decreased the rate of false positive species called in a sample relative to other NGS techniques and allows for greater sensitivity and taxonomic specificity. This assay may be especially useful for low colony-count or negative-culture samples to diagnose and guide patient treatment. Disclosures Mara Couto-Rodriguez, MS, Biotia Inc.: Employee of Biotia Inc. a for-profit biotechnology company David C. Danko, PhD, Biotia Inc.: Employee of Biotia Inc. a for-profit biotechnology company Xavier O. Jirau Serrano, MS, Biotia Inc.: Employee of Biotia Inc. a for-profit biotechnology company Taylor Paisie, MS, Biotia Inc.: Employee of Biotia Inc. a for-profit biotechnology company John C. Papciak, BS, Biotia Inc.: Employee of Biotia Inc. a for-profit biotechnology company Eszter Szollosi, BS, Biotia Inc.: Employee of Biotia Inc. a for-profit biotechnology company Christopher E. Mason, PhD, Biotia Inc.: Advisor/Consultant|Biotia Inc.: Board Member|Biotia Inc.: Ownership Interest Caitlin Otto, PhD, D(ABMM), Biotia Inc.: Advisor/Consultant Niamh B. O'Hara, PhD, Biotia Inc.: Board Member|Biotia Inc.: Ownership Interest Dorottya Nagy-Szakal, MD PhD, Biotia Inc.: Employee of Biotia Inc. a for-profit biotechnology company|Biotia Inc.: Stocks/Bonds.
Read full abstract