Rapid diagnosis of common, undetected, and uncultivable bloodstream infections from routine positive blood cultures | LC26

Abstract
Metagenomic sequencing could transform clinical microbiology by enabling rapid pathogen identification and antimicrobial resistance (AMR) prediction in critically ill patients with bloodstream infections (BSIs). However, clinical use has been limited by speed, accuracy, and technical feasibility. Our aim was to develop and evaluate a direct-from-positive blood culture workflow using Oxford Nanopore sequencing that overcomes these limitations, delivering results rapidly and accurately. We developed and evaluated a direct-from-positive blood culture metagenomic sequencing method for rapid pathogen and AMR prediction using Oxford Nanopore sequencing. Species prediction was performed using Kraken2 with a comprehensive standard database, employing heuristic and random forest classification models. Additionally, we benchmarked AMR classification tools and databases, including ResFinder, Comprehensive Antibiotic Resistance Database (CARD), and National Center for Biotechnology Information (NCBI) AMRFinderPlus. We processed 273 randomly selected blood cultures (211 positive and 62 negative) from hospitalised patients in real time, comparing species identification, AMR detection, and time-to-result against standard culture-based diagnostics performed by the hospital’s routine microbiology laboratory. Across all samples, our method achieved 97% sensitivity and 94% specificity for species identification compared with established diagnostic methods, improving to 100% for both after accounting for plausible additional infections. We detected 19 additional infections (13 polymicrobial, 5 previously unidentifiable, and 1 in a culture-negative sample) and delivered results within 3 h 20 min (interquartile range [IQR] 3 h 07 min to 3 h 27 min), nearly one third of the time required for routine methods. For the ten most common pathogens, AMR results were produced 20 h faster than conventional antimicrobial susceptibility testing, with 88% sensitivity and 93% specificity. Performance varied by species: for Staphylococcus aureus and Escherichia coli, AMR prediction sensitivity was 100% and 91%, and specificity was 99% and 94%, respectively. Metagenomic sequencing has the potential to rapidly and comprehensively detect pathogens and antimicrobial resistance in bloodstream infections. Integration into clinical practice could reduce diagnostic delays, limit empirical antibiotic use, and enable earlier targeted therapy. Further improvements in AMR prediction for specific species or drug classes, and multisite validation, are required before routine clinical implementation.

Biography
Dr Kumeren Govender, M.D., DPhil, is a physician, academic, and entrepreneur with extensive expertise in global health, medicine, and biotechnology. He served as the Chief Scientific Officer at the Ellison Institute of Technology, where he led strategic initiatives in pandemic preparedness and global health innovation, partnering with institutions such as Oxford University. Previously, Kumeren held leadership roles at the Global Pathogen Analysis Service and co-founded Neurolytic Healthcare, a precision medicine start-up raising over $3 million. A Rhodes Scholar, he earned a DPhil in Clinical Medicine at Oxford, specialising in metagenomic sequencing for clinical diagnostics. Kumeren’s accolades include the Oxford Vice-Chancellor’s Innovation Award, a Y-Combinator start-up fellowship, and recognition as one of South Africa’s Top 200 Young Achievers. He has authored numerous academic publications, and remains active in policy roles, including work with OpenAI on medical AI models.