Statistical analysis and classification of taxa associated with lung cancer

Statistical analysis and visualization, ANOVA and PCoA, was performed on sequencing quality metrics by population sample type (ImA, HB, NT, and T) (Additional file 1: Figure S1). Alpha- and beta-diversity metrics were computed in QIIME with differential abundance analyses performed in R and Stata (v13). Mann–Whitney tests corrected for multiple testing (Benjamini–Hochberg [FDR]) were used to conduct initial comparisons between tissue type and histological subtype (AD or SCC) followed by multivariable logistic regression controlling for multiple confounders (age, gender, race, smoking status, stage, antibiotic exposure, lung location, average Phred score, and sequencing run) (Additional file 1: Table S11). An additional logistic regression model was constructed to estimate the odds of AD versus SCC for each taxa separately (identified from the initial testing) stratified by TP53 mutation status (wild-type versus mutated) with and interaction term between the taxa and mutation added to the model. See Additional file 1: Supplemental Methods for details of statistical modeling.