Title: Machine Learning Holds Potential for Rapid Diagnosis of Lower Respiratory Tract Infections
(Data source: Preprints with The Lancet)*
In a recent study, researchers have explored the use of machine learning models combined with clinical information, metatranscriptomics, and lower respiratory tract microbiome to diagnose lower respiratory tract infections (LRTIs) more rapidly. Their findings suggest that these prediction models could serve as a valuable tool in circumventing the morbidity and mortality associated with traditional microbiological testing.
LRTIs are a major cause of global human mortality, responsible for over 3 million deaths annually. Conventional diagnostic methods for LRTIs have limitations, such as low sensitivity and the inability to identify 60-70% of causative agents. Moreover, these methods can take 24-48 hours or longer to characterize an infection.
The symptomatic presentation of LRTIs often overlaps with non-infectious conditions like asthma or chronic obstructive pulmonary disease (COPD), leading to challenges in accurate diagnosis. Delaying diagnosis or misdiagnosis can be life-threatening for patients.
Research in the field of microbial genomes has challenged the traditional view that the lungs are initially sterile. Instead, it suggests that LRTIs result from a combination of low microbial species diversity, high overall biomass, and host inflammation response.
Microbiome studies have also shown alterations in the respiratory tract microbiomes of non-infectious diseases like asthma, emphasizing their significance in LRTI identification and characterization. One emerging method, metagenomic next-generation sequencing (mNGS), has shown promise as a rapid and sensitive alternative to traditional diagnostic tools, providing accurate diagnoses within minutes to hours.
This preprint study aimed to combine respiratory microbiome and host transcriptional profiling with clinical data. Researchers enrolled 136 patients suspected of having LRTIs from the Peking University People’s Hospital between May 2020 and January 2021. The patients underwent traditional microbiological and serological testing for LRTI diagnosis, and bronchoalveolar lavage fluid (BALF) was collected for further analysis. BALF samples were subjected to DNA and RNA sequencing to identify the lung microbiome.
The analysis revealed 674 differentially expressed genes (DEGs), with 613 DEGs up-regulated in the LRTI cohort and 61 down-regulated compared to the non-LRTI group. The up-regulated DEGs associated with LRTIs were found to be linked to pathogen infection pathways according to the Kyoto Encyclopedia of Genes and Genomes (KEGG).
Using this data, the researchers trained a machine learning model using random forest algorithms, incorporating 11 clinical variables, 39 lung microbiome features, and 20 host response indicators. The model demonstrated an impressive diagnostic accuracy of 88.2%, with results available in just a few hours, a significant improvement over traditional diagnostic methods.
While this study represents a novel approach to LRTI diagnosis, it is important to note that metagenomic next-generation sequencing is currently expensive and requires high technical capabilities. Additionally, the machine learning models used in this study are diagnostic indicators and do not provide insights into the mechanisms or biological functions of microbiota-host transcriptome interactions.
If validated through peer review and further developed to reduce costs, this research could revolutionize the diagnosis of lower respiratory tract infections, enabling clinicians to rapidly and accurately identify infections, ultimately reducing the associated morbidity and mortality.
(Word count: 453)
*Important notice: SSRN publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.
Journal reference: Preliminary scientific report. Chen H, Qi T, Guo S, et al. (2023). Integrating Respiratory Microbiome and Host Immune Response Using Machine Learning for Diagnosis of the Lower Respiratory Tract Infections. Preprints with The Lancet. doi: 10.2139/ssrn.4505343 https://ssrn.com/abstract=4505343
