Multiomic profiling of circulating hepatocyte-derived extracellular vesicles for hepatocellular carcinoma detection in liver cirrhosis

Abstract

Motivation: Liver cirrhosis is the primary risk factor for hepatocellular carcinoma (HCC), with current surveillance tests often failing to detect the disease during its curative stages. Extracellular vesicles (EVs) can carry multiple types of biological molecules, from diseased tissues such as tumors, and have gained traction recently as a promising source for biomarker discovery in liquid biopsy. Results: In this multicenter case-control discovery cohort, we targeted and captured hepatocyte-derived extracellular vesicles from patients’ plasma samples to identify omics biomarkers for detecting HCC across all stages, with a particular focus on early-stage cases. By leveraging recent advances in mass spectrometry and small-RNA sequencing to overcome common challenges in blood-derived EVs research, such as the low yield of biomolecules, we obtained results indicating that our targeted method effectively captures EVs involved in liver-related pathways. Using recursive feature elimination (RFE) and random-forest, a preliminary proteomic signature derived from a training dataset achieved an Area Under the Curve (AUC) of 0.74 in unseen test samples. A final multiomics panel, composed of six proteins identified through mass spectrometry, four microRNAs from small RNA-sequencing, and alpha-fetoprotein (AFP) measured by ELISA demonstrated robust generalizability across a dataset composed of samples from different cancer stages and cirrhosis aetiologies. The multiomic biomarker discovery strategy was further validated using RFE and random-forest models with random partitioning into training and testing sets, yielding promising performance results when applied to test subsets with HCC samples across all stages (mean-AUC: 0.85, mean-sensitivity: 0.8, mean-specificity: 0.81), as well as testing subsets composed solely of early-stage cases (mean-AUC: 0.81, mean-sensitivity: 0.67, mean-specificity: 0.81). These findings underscore the potential of using multiomics analysis of circulating hepatocyte-derived EVs to improve early detection of HCC in cirrhotic patients.