The opposite pattern was observed for MLR model based on female specific metabolites, which resulted in a good predictive power (AUC = 0.87, p < 0.0001) for RCC diagnosis among females ( Figure 5 C), while in males this model was borderline significant with p = 0.02 and AUC of 0.68 ( Figure 5 D).
Importantly, when the same set of metabolites was used in female cohort, the MLR model did not reach statistical significance ( p = 0.0575), indicating that indeed these three metabolites were only associated with RCC status in male subjects ( Figure 5 B) and showed poor performance in females.
Contrary to lactate, pyruvate levels were increased in RCC patients ( Figure 6 C) and its interaction with gender almost reached statistical significance ( p = 0.07), which was consistent with pyruvate being selected by female specific PLS-DA model ( Figure 4 D).