Serological proteomic characterization for monitoring liver fibrosis regression in chronic hepatitis B patients on treatment

Longitudinal serological proteomic dynamics during antiviral therapy (AVT) in chronic hepatitis B (CHB) patients with liver fibrosis remain poorly characterized. Here, using four-dimensional data-independent acquisition mass spectrometry (4D-DIA-MS), paired liver biopsy (LBx)-proven serum samples from 130 CHB liver fibrosis patients undergoing short-term (78 weeks) or long-term (260 weeks) AVT are analyzed. Our findings show that prolonged AVT drives progressive serological proteomic remodeling in fibrosis regressors, characterized by a temporal inversion in the activation of the complement and coagulation cascades. Using machine learning algorithms trained on the 4D-DIA-MS discovery cohort, we develop a logistic regression model incorporating a seven-protein panel for short-term AVT and a three-protein panel for long-term AVT, respectively, both of which demonstrate moderate discriminatory capabilities for fibrosis regression. Subsequent external validation in an independent cohort (n = 54) with serial LBx assessments at baseline, 78 weeks, and 260 weeks, where serological proteins are quantified using parallel reaction monitoring mass spectrometry (PRM-MS), further confirms their generalizability. Furthermore, our longitudinal trajectory analysis highlights that the long-term proteomic signature exhibits greater stability compared to the short-term panel. This study proposes and validates duration-adapted serological proteomic panels as non-invasive tools for monitoring histological fibrosis regression in on-treatment CHB patients. Liver fibrosis regression during antiviral therapy in chronic hepatitis B patients is not well understood. Here, the authors use mass spectrometry and machine learning to develop and validate serological proteomic panels, offering non-invasive tools for monitoring fibrosis regression.