Fig. 3: Non-invasive diagnostic tool to identify patients with at-risk NASH.
From: A proteo-transcriptomic map of non-alcoholic fatty liver disease signatures
a, Binary logistic regression modelling identified a composite model in the discovery cohort (n = 191 patients) that could classify patients with at-risk NASH on the basis of the variables BMI, T2DM and circulating ADAMTSL2, AKR1B10, CFHR4 and TREM2. b, Performance of the classification model in two independent cohorts (discovery n = 191 patients, validation n = 115 patients) in comparison with NFS, Fibrosis-4 (FIB-4) and AST:ALT ratio. Bar charts present AUC for each score with the corresponding standard error of area, as calculated by ROC analysis. Paired-sample area difference under the ROC curve test was used to compare the classification model with the other scores (discovery cohort NFS P = 0.000001, FIB-4 P = 7.8667 × 10−7 and AST:ALT ratio P = 6.85 × 10−10; validation cohort NFS P = 0.000429, FIB-4 P = 0.000949 and AST:ALT ratio P = 0.000036) (****P < 0.0001, ***P < 0.001). c, Representative images of immunohistochemistry and quantification for AKR1B10 in human liver biopsies (n = 30 biologically independent patient samples). Scale bars, 100 μm. Data are presented as mean ± s.e.m. (Kruskal–Wallis with Bonferroni correction and Mann–Whitney U-test). Arrows indicate necro-inflammatory region with ballooned hepatocytes.
