Deep learning models reading clinical data and liver omics strongly distinguish NASH from steatosis and suggest new genes involved in liver disease severity

Background & Aims Metabolic dysfunction-associated steatotic liver disease (MASLD) is a frequent co-morbidity of obesity and diabetes, with prevalence increasing worldwide. Recognising liver disease stages and elucidating the molecular underpinning of their progression are thus medically important. Methods Using data gathered from 300 patients with obesity of the ABOS cohort, we selected non-redundant clinical variables, gene expressions and CpGs methylation levels most associated with severity using unsupervised approaches to train a multi-module, multi-layer perceptron predicting patients liver status. Results The combination of five models trained on the three modalities reached an AUC of 0.945 on a validation set, with accuracies above 81% for simple steatosis and 88% for NASH, outperforming other machine learning models so far. The latent space of our clinical data module retrieved patient clusters identified previously, but not the gene expression and DNA methylation ones. While all three modules are needed to reach the best prediction accuracy in all classes, the gene expression module impacted the decision the most. Independently trained models put more weights on the same genes, underlying their importance. Impactful genes were linked to immune responses and extracellular matrix. Many of those genes were previously unassociated with MASLD onset or progression. Conclusions A multi-omics deep-learning model can recognise steatohepatitis from simple liver steatosis with an AUC of 0.945 and identify new genes potentially involved in NAFLD progression. Gene expressions profiles predicting disease severity are largely different from those specific of clinical variable clusters. Impact and implications: This study suggests that clinical variables are not sufficient to recognise the severity of steatotic liver disease with high accuracy, but model efficiency increases when used together with liver epigenetics and transcriptomics.