A multicentre evaluation of metagenomic sequencing for pathogen detection in central nervous system infections

Metagenomic next-generation sequencing (mNGS) is a promising tool for diagnosing central nervous system infections. However, the low-biomass nature of cerebrospinal fluid (CSF) increases susceptibility to contamination and host-background interference, potentially compromising accuracy. This study aimed to evaluate CSF mNGS performance across multiple laboratories and to identify key factors influencing detection accuracy.

A reference panel of 15 CSF samples was designed to evaluate CSF mNGS performance across laboratories, including 3 replicate samples, 5 serial concentration-gradient samples, 3 anti-interference samples with added human serum albumin or increased host nucleic acids, and 3 simulated clinical case samples, along with 1 negative sample. A total of 127 laboratories participated, which apply mNGS in clinical diagnostics or research. Each laboratory used independently developed mNGS workflow, which varied in experimental procedures, bioinformatic pipelines, and positive detection thresholds. Accuracy, repeatability, sensitivity, and anti-interference capability were systematically evaluated, and sources of erroneous results and methodological factors influencing accuracy were analysed.

Overall performance across 127 laboratories was favourable (average F1-score 0.98, reflecting overall accuracy by balancing sensitivity and specificity). Most false-positive results (83.43%) were due to experimental contamination, whereas false negatives were mainly attributed to RNA viruses (57.14%). Methodological factors significantly affected detection, with impact varying by microbial type. Generally, pelleting impaired the detection of all microbes. Notably, microbial enrichment through DNase treatment and Kraken2 improved detection accuracy for DNA viruses, bacteria, fungi and atypical pathogens, but had little effect on RNA viruses.

This large-scale study underscores the need for improved contamination controls, optimized RNA virus detection, and enhancement of key wet-lab procedures to strengthen CSF mNGS reliability. These findings provide actionable insights to refine mNGS workflows and advance its clinical utility for diagnosing central nervous system infections.