Fig. 5: ACAs improve prediction of T1D.
From: Use of a glycomics array to establish the anti-carbohydrate antibody repertoire in type 1 diabetes
a Receiver-operator characteristic curve for three predictive models of T1D. A null model including sex, HLA risk, FDR, and draw age (black). A full model including all variables in the null model plus the first principal component from each of the 11 clusters of ACA (red). A partial model including all variables in the null model plus the first principal component from each of the 5 clusters of ACA significantly associated with progressors or non-progressors (blue). b The variance accounted for by each component of the full model from A. Partial correlation coefficient (R2) values from full model was plotted (black bars) against the variables (x-axis). c The area under the receiver-operator characteristic curve for a random number of glycans included in a ridge regression model with clinical variables from 1 to 200. AUC (dots, y-axis) values were plotted against number of glycans (x-axis), the dots represent mean AUC value while error bars represent 95% confidence intervals for the AUC values. Source data are provided as a Source Data file.
