To develop a multimodal machine learning-based diagnostic approach for differentiating lung cancer from pulmonary infections using bronchoalveolar lavage fluid (BALF) metagenomic next-generation sequencing (mNGS) data, highlighting its significance in clinical diagnostics.
Key Findings:
Model VI achieved an AUC of 0.937 in the training cohort and 0.847 in the test cohort, indicating strong diagnostic performance.
Rule-in/rule-out strategy improved accuracy to 0.896 for tuberculosis, 0.915 for fungal, and 0.907 for bacterial infections, demonstrating clinical relevance.
Distinct microbial profiles and host responses were identified between lung cancer and pulmonary infections, suggesting unique diagnostic markers.
Interpretation:
The findings suggest that mNGS-based multimodal analysis can serve as a cost-effective and rapid diagnostic tool for distinguishing lung cancer from pulmonary infections, potentially transforming clinical decision-making.
Limitations:
Study limited to a single center and specific patient demographics, which may affect generalizability.
Further validation in diverse populations and settings is needed to confirm findings.
Conclusion:
mNGS represents a promising approach for early and accurate differential diagnosis of lung cancer and pulmonary infections, potentially improving patient management and outcomes.
