Fig. 1: Study flow chart, machine learning algorithms and their performance when using the three prediction models. | npj Digital Medicine

Fig. 1: Study flow chart, machine learning algorithms and their performance when using the three prediction models.

From: Discovering a trans-omics biomarker signature that predisposes high risk diabetic patients to diabetic kidney disease

Fig. 1

a The scheme illustrates the data processing and machine learning workflow that integrates the non-targeted metabolites, lipidomics (P180-metabolites), SNPs, and clinical data. b The two stage modeling workflow used to predict diabetes mellitus (DM) and chronic kidney disease (CKD). c The Confusion Matrix of prediction accuracy obtained by aggregating the three models into four groups. For example, the Non DM and Non CKD predicted label is predicted as Non DM in model 1 and predicted to be Non CKD in model 3. d–e The receiver operating characteristic (ROC) curves of Model 1, Model 2 and Model 3 that were used for predicting DM and CKD in the training cohort (d) and the validation cohort (e). Abbreviations include DM diabetes, CKD chronic kidney diseases, DKD diabetic kidney diseases, AUC area under the curve, ACC accuracy, SVM Support Vector Machine. The figure was created with BioRender.com.

Back to article page