Figure 2

Distribution and area under curve for oral-glucose tolerance test of insulin and glucose, as well as feature importance for constructed variables based on extreme gradient boosting models and linear regression. Panel (A)–(B) shows distributions for insulin and glucose related OGTT variables that were used to generate AUC variables for insulin and glucose. Panel C shows association between AUC insulin and AUC glucose, along with correlation coefficient. In Panel (D)–(E), relative importance for predictors generated by extreme gradient boosting models, using pre-processing techniques for metabolomics data to reduce number of predictors in the final model. Model diagnostics (RMSE) and validation (R²) are presented next to each prediction model. The most important predictors identified through prediction modeling were included in a linear regression model. Significance level are described as follows: *p-value < 0.05, p-values < 0.06. < denotes that lower levels for the predictor was associated with the target variable. > denotes that higher levels for the predictor was associated with the target variable.