Multivariate data analyses

Multivariate analyses were done with unsupervised principal component analysis (PCA) and supervised orthogonal partial least square discriminant analysis (OPLS‐DA) using SIMCA v. 14 (Umetrics, Umeå, Sweden). By including parameters like HbA_1c(%), BG, AER, body weight (BW) and kidney weight (KW) (final measuring time point) in the OPLS‐DA model as a y‐variable, an examination can be performed of whether the parameter drives the separation of the samples or not. In this study the four groups of mice (db/db insulin, db/db liraglutide, db/db vehicle and healthy control) were defined as separate classes in the OPLS‐DA analyses. The Variable Importance for the Projection (VIP) plots visualizes variables separating the classes (reflecting the latent structures), where levels >1 indicate influence on group separation (Xie et al. 2015). In the shared and unique structures (SUS) plot three groups are compared simultaneously based on their covariance and correlation (Wiklund 2008; Wiklund et al. 2008). The SUS plot was applied on the db/db insulin and db/db liraglutide groups in relation to the db/db vehicle mouse group (common control) and unique and shared structures were visualized.

In all multivariate analyses, the variables albumin and β‐globin were removed, since their intensities were identified to be an artifact from two of the samples. All the samples, including technical replicates, were examined individually and thereafter the two technical replicates were joined for the statistical analyses.