Abstract
Background: Accurate diagnosis of chronic Hepatitis B virus (HBV) infection related diseases is crucial to guide the therapy and to understand the mechanisms of disease progression. Plasma microRNAs, as stable biomarkers, have drawn significant attentions for distinguishing HBVrelated diseases.
Methods: In this study, a new HBV-related disease identification method based on a two-layer logistic regression model was presented. A total of nine effective plasma microRNA biomarkers were selected through sample collection, data processing, model selection, feature selection and model optimization to distinguish HBV-related chronic hepatitis and cirrhosis samples as well as healthy controls. The first layer utilized three microRNAs to distinguish HBV-related disease samples from healthy controls. Then the second layer divided the HBV-related disease samples into cirrhosis and chronic hepatitis samples by using eight microRNAs. Result: Test on two independent cohorts showed high accuracy and robustness of our model. Functional analysis of the selected microRNAs and their target genes confirmed that they were significantly associated with HBV-related diseases and related functional pathways. Conclusion: Compared with previous models, the two-layer model was more consistent with the underlying pathological progress of HBV related diseases from health to chronic hepatitis and further to liver cirrhosis. It could also take the results of other diagnostic tests into account, which could be useful in both physical examination and disease diagnosis.Keywords: Two-layer model, logistic regression, HBV, chronic hepatitis, cirrhosis.
Related Journals
Current Computer-Aided Drug Design
Article Metrics
20
1
1
1