Ecological heterogeneity beneath the pneumonia label: pathogen community gradients in clinical targeted sequencing of bronchoalveolar lavage fluid

Background A clinical diagnosis of pneumonia encompasses infections with fundamentally different microbial ecologies. We asked whether pathogen community analysis of tNGS data from bronchoalveolar lavage fluid (BALF) could reveal ecological heterogeneity masked by the binary pneumonia label. Methods We analyzed 1,247 BALF tNGS samples (December 2025–July 2026). Read counts were RPM-normalized and residualized against library size. Community differences were assessed by PERMANOVA (Bray-Curtis) with PERMDISP validation. Principal component analysis (PCA) identified ecological axes; an a-priori dysbiosis index (literature-defined nosocomial minus oral pathogens) provided convergent validation. Indicator species analysis (IndVal) identified pathogens driving community differences. A dual-axis quadrant framework (PC1 × PC2) was defined for clinical translation. Results Pneumonia status alone explained a statistically significant but ecologically negligible fraction of community variance (PERMANOVA on Bray–Curtis distances: pseudo-F=2.77, R 2 = 0.23%, p=0.001). PCA identified two orthogonal, library-size-independent axes: PC1 (5.4%, oral enrichment) and PC2 (4.1%, nosocomial burden). PC2 correlated with respiratory failure (r=0.19, p<0.001), sepsis (r=0.18, p<0.001), and stroke (r=0.21, p<0.001). The a-priori dysbiosis index converged with the gradient (pneumonia Q1→Q5: 37%→50%, respiratory failure 6%→21%). Four ecological quadrants defined by axis position captured graded severity: the two nosocomial-high quadrants (Type III and IV, 20% each) showed higher respiratory failure (17.2% and 14.2%) and sepsis (9.2% and 7.1%) than the nosocomial-low quadrants (Type I, 29%: 7.8% and 2.5%; Type II, 31%: 10.4% and 4.1%). Temporal validation (early vs late cohort) and sensitivity analyses confirmed gradient stability. Conclusions The pneumonia label aggregates ecologically distinct communities. Pathogen community analysis extracts clinical information from tNGS data that binary diagnosis misses. A dual-axis ecological framework supplements the pneumonia label by positioning each sample along an oral-to-nosocomial gradient that tracks with clinical severity.