Nanopore-based metagenomic methods have proven useful for identifying bacterial causes of infections. Yet, thorough assessments of their effectiveness in everyday clinical practice for simultaneously detecting bacteria, fungi, and viruses remain limited. We created a paired sequencing strategy called the Nanopore Meta-Panel process (NanoMP). For each respiratory sample, it ran an untargeted metagenomic workflow (Meta) alongside a targeted enrichment workflow (Panel). This prospective multicenter trial included 450 respiratory samples. A machine learning-driven filtering framework, refined with ROC analysis from orthogonal testing on 21 key species, enhanced detection reliability. AMR gene detection relied on the CARD database combined with SNP profiling. In the Meta workflow, sensitivity reached 82.9% for bacteria, 88.7% for fungi (excluding Aspergillus species), and 75.0% for the Mycobacterium tuberculosis complex. The Panel workflow, through targeted amplification, substantially raised sensitivity for viruses (over 80.0% versus 39.4%) and Aspergillus (81.8% versus 42.3%). Across both processes, NanoMP delivered 80.2% sensitivity and 98.8% specificity against the composite reference standard. It also reliably pinpointed key AMR determinants like blaKPC-2, blaOXA-23, and mecA, correctly linking them to their host organisms even in polymicrobial cases. The NanoMP strategy merges unbiased metagenomics with targeted enrichment on the same specimen via parallel Nanopore runs. This enabled comprehensive detection of bacterial, viral, and fungal pathogens in lower respiratory samples and showed strong real-world clinical utility.
