Metabolic Profiling and Early Diagnosis of Alcoholic Fatty Liver Disease Using Support Vector Machine Model

Abstract

Alcoholic fatty liver disease (AFLD) is one of the most common pathological changes associated with alcoholic liver disease. This study is aimed at investigating the specific metabolic changes occurring in AFLD and to develop a diagnostic method based on the blood metabolomics of AFLD. Twenty-four rats were randomly divided into an AFLD group and a control group. The AFLD model was established by administering 40% alcohol and verified by pathologic examination. Metabolic changes in the blood were investigated by GC-MS, and both hepatic function and metabolic ability were assessed. Using the metabolic data, a diagnostic model was developed with a support vector machine (SVM). The model was validated using cross-validation techniques and achieved a classification accuracy of 100%. Additionally, there were statistically significant differences in metabolic, hepatic function, and pharmacokinetic changes between the two groups. The level of urea, hydroxysuccinic acid, 2-propenoic acid, total cholesterol, and high-density lipoprotein cholesterol increased, area under the concentration-time curve (AUC) (0−t), AUC (0−∞), and Cmax of phenacetin shorten in the AFLD group (p<0.05). The classification accuracy of the SVM model based on metabolic data was 100%. In conclusion, the metabolic ability was heightened in the early stage of AFLD, leading to accelerated metabolism of urea and hydroxysuccinic acid. The SVM model can be used to detect early changes in AFLD based on metabolic data.