Background
Given the increasing prevalence of metabolic dysfunction-associated fatty liver disease (MAFLD) and non-alcoholic steatohepatitis (NASH), there is a critical need for accurate non-invasive early diagnostic markers.
Objective
This study aimed to validate NLRP3-related RNA signatures (EP300, CPN60, and ITGB1 mRNAs, miR-6881-5p, and LncRNA-RABGAP1L-DT-206) using an integrated molecular approach and advanced machine-learning algorithms to identify robust biomarkers for early diagnosis of NASH.
Methods
A cohort of 237 participants (117 Healthy controls, 60 MAFLD, 120 NASH) was utilized. Twenty-five demographic, clinical, and molecular features were collected from each participant. Various machine learning models were trained on the dataset.
Results
The Random Forest algorithm emerged as the most effective classifier. The model identified nine key features: EP300 mRNA, CPN60 mRNA, AST, D. bilirubin, Albumin, GGT, HbA1c, HOMA-IR, and BMI, achieving an impressive 97 % accuracy in distinguishing NASH from non-NASH cases.Conclusion
The integration of molecular, clinical, and demographic data with machine learning algorithms provides a highly accurate method for the early diagnosis of NASH. This model holds promise for early detection in individuals at risk of progressing to cirrhosis or liver cancer and may aid in identifying new therapeutic targets for managing NASH.