Non-alcoholic fatty liver disease (NAFLD) encompasses a broad spectrum of conditions from simple steatosis (non-alcoholic fatty liver (NAFL)) to non-alcoholic steatohepatitis (NASH), and its global prevalence continues to rise. NASH, the progressive form of NAFLD, has higher risks of liver and non-liver related adverse outcomes compared with those patients with NAFL alone. Therefore, the present study aimed to explore the mechanisms in the progression of NAFLD and to develop a model to diagnose NASH based on the transcriptome and epigenome. Differentially expressed genes (DEGs) and differentially methylated genes (DMGs) among the three groups (normal, NAFL, and NASH) were identified, and the functional analysis revealed that the development of NAFLD was primarily related to the oxidoreductase-related activity, PPAR signaling pathway, tight junction, and pathogenic Escherichia coli infection. The logistic regression (LR) model, consisting of ApoF, THOP1, and BICC1, outperformed the other five models. With the highest AUC (0.8819, 95%CI: 0.8128-0.9511) and a sensitivity of 97.87%, as well as a specificity of 64.71%, the LR model was determined as the diagnostic model, which can differentiate NASH from NAFL. In conclusion, several potential mechanisms were screened out based on the transcriptome and epigenome, and a diagnostic model was built to help patient stratification for NAFLD populations.
Keywords: DNA methylation; logistic regression; machine learning; non-alcoholic fatty liver disease; non-alcoholic steatohepatitis.