Fig. 2: Microbial community composition is also a determinant of metabolite profiles.

Univariate analysis of nicotine-free and nicotine plus aerosol using partial least squares discriminant analysis (PLSDA) demonstrated that 39.1% of the variation between biofilms was explained by component 1 following exposure to nicotine-free ENDS (Q² = 0.75525, R² = 0.80226), while exposure to nicotine-containing vapor accounted for 37.6% of the variation of component 1 (Q² = 0. 73132, R² = 0. 79832, A, B). indicating that biofilm diversity was also a robust determinant of metabolite profile. A machine-learning algorithm trained on the metabolite profiles identified 15 compounds as discriminants of biofilm composition and vapor type (C). Covariate analysis using two-way ANOVA demonstrated a significant difference in 13 metabolites based on biofilm exposure alone, 17 metabolites for ENDS exposure alone, and 17 metabolites for the interaction between biofilm and ENDS exposures (p < 0.05, ANOVA, D). All samples were run in triplicates, and the experiments were duplicated. Only peaks that exceeded the zero threshold were inducted into analysis. Data supporting this figure can be found in Supplemental File 2.