Figure 5 | Scientific Reports

Figure 5

From: Predicting type 2 diabetes via machine learning integration of multiple omics from human pancreatic islets

Figure 5

DIABLO prediction. (a) Comparison of type 2 diabetes (T2D) prediction accuracy of the integrative multiOmics DIABLO model (component 2) with the predictive capacities of the selected features of respective individual Omics (mRNA expression (expr), DNA methylation (meth), genetic variation/SNPs (gen)) as well as clinical phenotypes (phen). Sig. *: Mann-Whittney U test p = 0.02, Sig. ***: Mann-Whittney U test p = 0.00017. (b) Comparison of T2D prediction in the form of receiver operating characteristic (ROC) curves for the integrative multiOmics DIABLO and individual Omics models (mRNA expression (expr), DNA methylation (meth), genetic variation/SNPs (gen)) as well as clinical phenotypes (phen) (PLS component 1). The ROC area under the curve (AUC) for the multiOmics DIABLO model was significantly higher than both the ROC AUC for expression and the ROC AUC for DNA methylation with Mann-Whittney U test p = 0.01.

Back to article page