Extended Data Fig. 7: Relationship between the reduced representation of the glucose time-series and the extracted curve features. | Nature Biomedical Engineering

Extended Data Fig. 7: Relationship between the reduced representation of the glucose time-series and the extracted curve features.

From: Prediction of metabolic subphenotypes of type 2 diabetes via continuous glucose monitoring and machine learning

Extended Data Fig. 7

We calculated the correlations between the top 2 PCs (OGTT_G_ReducedRep) and the extracted features (OGTT_G_Features). Variances explained by PC1 and PC2 are 43.8% and 18%, respectively. As illustrated in the Figure, PC1 strongly negatively correlated with the incremental area under the curve (iAUC), and positively correlated with the negative area under the glucose curve (nAUC). This means, for example, that higher positive values on the PC1 are associated with lower iAUC.

Back to article page