Here, we will leverage longitudinal multi-omics data of 3D primary human liver cultures from NASH patients and controls developed in the hosting lab. These comprise proteomics, transcriptomics, lipidomics and phosphoproteomics data sets from fatty liver samples with steatosis, insulin resistance and fibrosis, regenerating livers after injury as well as healthy controls. Here, we aim to integrate these longitudinal data with comprehensive omics datasets of NASH cohorts to construct and train a machine learning classification model with good predictive performance, followed by identifying key genes associated with NASH on this basis utilizing approaches to interpreting model prediction. These analysis aspire to provide the following deliverables: 1) Establish a robust NASH disease classification model based on omics data; 2) Identify a set of interpretable benchmark key genes for NASH; 3) Integrate multi-omics data into a unified analytical framework. Combined, this project will deepen our knowledge of NASH with an interpretable approach aligning state-of-the-art computational research with mechanistic discovery and clear translational benfits.