Fig 2: ROC curve graph between each group obtained through machine learning. (A) ROC curve for the combination of RBP4 and NOX4 between normal and MCI using the SVM algorithm. AUC is shown as 0.83. (B) ROC curve for the combination of RBP4, CXCL10, and NOX4 between normal and SCI using the Extra Tree algorithm. AUC is shown as 0.78. (C) ROC curve for the combination of RBP4 and NOX4 between normal and CCDS using the SVM algorithm. AUC is shown as 0.75.

Fig 3: The correlation heatmap manifested the correlation between five variables: RBP4, CXCL10, NOX4, CCDR, and age. The color of each cell indicates a high correlation in red and a low correlation in blue, with the darker color indicating the strength of the correlation.

Fig 4: Protein profiling of the serum from APP/PS1 mice and MPTP-induced mice for disease model animals was obtained by performing proteome array analysis. (A) Images of array spots on the membrane are proteome array for immune blot, including 111 cytokine, chemokine, and growth factor types. The red rectangle indicates RBP4 and CXCL10 showing expression levels in the APP/PS1 and MPTP-induced mice. Magnifying indicates an enlarged image of RBP4 and CXCL10 in the square below each image. Array spots were analyzed according to the manufacturer’s instructions. (B) Quantifying RBP4 and CXCL10 levels in control and disease model animals from serum. Data are mean ± standard error of the mean (SEM). *, p < 0.05; **, p < 0.01; ***, p < 0.001 vs. control by a Student’s two-tailed t-test. AD: Alzheimer’s disease; PD: Parkinson’s disease.

Fig 5: RBP4, CXCL10, and NOX4 protein levels in dogs of normal and CCDS groups. (A) Representative immunoblot analysis of RBP4, CXCL10, and NOX4 expression in CCDS groups compared to normal groups. (B) Quantifying RBP4, CXCL10, and NOX4 levels in normal and CCDS groups from plasma. Data are mean ± SEM. *, p < 0.05 vs. normal by a Student’s two-tailed t-test.