Fig. 5

Diagnostic potential of oral microbial markers for SLE. A Two bacterial markers were selected as the best markers set by random forest model through five-fold cross-validation. B Importance distribution map of the selected microbial markers in the model. C The POD value was significantly higher in SLE (n = 100) compared with that in HC (n = 200) in the derivation cohort. D The POD value achieved an AUC of 0.9166 (95% CI 0.8848–0.9483) between SLE (n = 100) versus HC (n = 200) in the derivation cohort (p < 0.0001). E The POD value was significantly higher in SLE (n = 40) than that in HC (n = 80) in the independent validation cohort (p < 0.05). F The POD value achieved an AUC of 0.8422 (95% CI 0.7687–0.9157) between SLE (n = 40) versus HC (n = 80) in the independent validation cohort (p < 0.0001). G The POD value was significantly higher in SLE (n = 42) than that in HC (n = 80) in the cross-reginal validation cohort (p < 0.05). H The POD value achieved an AUC of 0.8406 (95% CI 0.7677–0.9135) between SLE (n = 42) versus HC (n = 80) in the cross-reginal validation cohort (p < 0.0001). CV Error, cross-validation error; HC healthy controls, SLE systemic lupus erythematosus; POD probability of disease, OUT operational taxonomic unit, CI confidence interval, AUC area under the curve