Fig. 2

Microbial taxa related to survival in tumor and normal lung from stage II NSCLC patients. a For taxa selected >125 times in 500 × 10-fold cross-validated elastic-net penalized Cox regression and with FDR-adjusted q < 0.20 in either the tumor (n = 39) or normal lung (n = 41) data, we show number of times selected out of 500 times, and the hazard ratio (95% CI) from Cox proportional hazards regression of RFS, DFS, or OS on clr-transformed taxon abundance, adjusted for age, sex, race, histology, smoking status, and chemotherapy. b Heatmap of clr-transformed taxon abundance in normal lung tissue, sorted by months of RFS. c Boxplots of clr-transformed taxon abundance in normal lung tissue according to 1-year RFS status. P-values from Wilcoxon rank-sum test. d Comparison of study results with those from Peters et al. [21] and Tsay et al. [17]. Plots show the hazard ratio (95% CI) from Cox proportional hazards regression of OS on clr-transformed taxon abundance. Peters et al. (2019) included 16 tumor and 17 normal lung tissue samples, and models were adjusted for age and sex. Tsay et al. [17] included lower airway brushings from local stage (I-IIIA) NSCLC patients (n = 34 patients, n = 93 samples [36 involved, 57 non-involved]), and models were adjusted for age, sex, race, histology, smoking status, chemotherapy, and surgery; models also accounted for clustering of samples within patients