ORIGINAL RESEARCH article
Front. Cell Dev. Biol.
Sec. Cellular Biochemistry
This article is part of the Research TopicAdvanced AI and Omics Integration for Liver Disease ResearchView all articles
A Machine Learning-Based Predictive Model for 48-Week Hepatitis B Surface Antigen Seroclearance in Chronic Hepatitis B Patients Treated with Pegylated Interferon α-2b: Prediction at Week 24
Provisionally accepted
- Shanghai East Hospital, School of Medicine, Tongji University, Shanghai, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Background: Chronic hepatitis B (CHB) is an infectious disease mainly affecting the liver, caused by the hepatitis B virus (HBV). In the treatment of CHB, pegylated interferon α-2b (PEG-IFNα-2b) is one of the important therapeutic options. However, there are significant individual differences in patients' responses to this treatment and only a few patients can achieve hepatitis B surface antigen (HBsAg) seroclearance. Therefore, an effective method to identify patients with a high likelihood of favorable response at an early stage is urgently needed. Methods: In this study, we analyzed data from CHB patients who received antiviral treatment with PEG-IFNα-2b and completed 48 weeks of follow-up in the "OASIS" Project. Patients were divided into the seroclearance group and the non-seroclearance group based on whether HBsAg seroclearance was achieved at week 48.Five distinct machine learning feature selection algorithms were used to identify the optimal predictive variables for HBsAg seroclearance. These key variables were then incorporated into 12 machine learning algorithms to build predictive models for HBsAg seroclearance. The best-performing model was selected, and its performance was evaluated. Results: A total of 680 subjects were included in this study, comprising 165 in the 48-week seroclearance group and 515 in the 48-week non-seroclearance group. Through 5 different machine learning feature selection algorithms, 11 variables were identified and used to construct 12 distinct machine learning models. Comparative analysis of these models, based on the Area Under the Receiver Operating Characteristic Curve (AUC) and Decision Curve Analysis (DCA) results from the training set, indicated that the Random Forest model was the optimal model for predicting HBsAg seroclearance. Conclusion:The Random Forest model effectively predicted the 48-week HBsAg seroclearance rate using indicators measured at 24 weeks of PEG-IFNα-2b therapy. This model can provide a reliable reference for optimizing clinical treatment strategies.
Keywords: Chronic hepatitis B, HBsAg seroclearance, predictive model, Machinelearning, clinical utility
Received: 29 Oct 2025; Accepted: 17 Nov 2025.
Copyright: © 2025 Kong, Wang, WANG, Lou, ZHOU, Wang, TAN and QU. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Lihong QU, 1905365@tongji.edu.cn
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.