# Metagenomic next-generation sequencing of alveolar lavage fluid improves the detection of pulmonary infection

## TL;DR

A study compared two metagenomic sequencing methods for detecting lung infections, finding that one method was more reliable for accurate pathogen detection.

## Contribution

The study compares two mNGS methods (Q and PACE) for pulmonary infection detection, highlighting their performance differences in clinical settings.

## Key findings

- Mycobacterium tuberculosis was predominantly detected via general mNGS.
- PACE showed high sensitivity and accuracy, while Q maintained 100% sensitivity but lower specificity.
- Bacterial-viral co-infections were the most common type identified.

## Abstract

This study evaluated the effectiveness of metagenomic next-generation sequencing (mNGS) in detecting pathogens in patients with pulmonary infections, comparing a low-data-volume, human-depleted quantitative (Q) method and a high-data-volume, non-human-depleted pathogen capture engine (PACE) method. A total of 133 patients were enrolled, comprising 59 in a control group (traditional culture) and 74 in an mNGS group (51 Q and 23 PACE). Bronchoalveolar lavage fluid samples were collected for pathogen detection. Mycobacterium tuberculosis was predominantly detected via general mNGS, whereas Candida albicans and Epstein-Barr virus were more frequently identified by PACE and Q, respectively. Among participants, 22.97% had bacterial mono-infections, and 2.70% had viral mono-infections; the most common co-infection involved bacteria and viruses (25.68%). Patients with fever, abnormal white blood cell, neutrophil percentage, and D-dimer levels exhibited higher detection rates. PACE showed consistently high sensitivity (decreasing from 100 to 92% as thresholds became more stringent) and specificity and accuracy that peaked at 100 and 96%, respectively. The Q method maintained 100% sensitivity at the lowest threshold but showed variable specificity (0.52–0.67) and accuracy (71–75%). These findings highlight the need for caution in clinical applications when using low-data-volume, human-depleted approaches, especially for complex pulmonary infection cases.