Identification and evaluation of biomarkers for diagnosis of chronic hepatitis B using RNA-seq

Chronic hepatitis B (CHB) is a global public health problem affecting hundreds of millions of people and is associated with significant morbidity and mortality of liver cancer. Exosomes originate from cells and their detection in biofluids provides valuable insights into cellular and tissue alterations, thus reflecting underlying pathological states. The aim of this study was to provide exosomal RNA biomarkers of CHB and develop a machine learning model for the non-invasive diagnosis of CHB patients.

The differentially expressed genes (DEGs) were screened according to the RNA-seq data of normal and CHB liver tissues. The biomarkers were selected according to the analysis of pathway enrichment and functional annotation. The correlation of biomarkers' expression level with the inflammation stage of CHB patients was analyzed. The non-invasive diagnostic value of the potential RNA biomarkers was evaluated by checking their different expression level in the plasma exosome of healthy individuals and CHB patients. A machine learning model was constructed to diagnose CHB by combining three identified biomarkers.

A total of 1,006 differential expressed genes (569 upregulated and 437 downregulated) were screened between normal and CHB tissues. The GO and KEGG results showed the DEGs were mainly enriched in inflammation-related pathways. Among these genes, the expression of 4 upregulated genes and 27 downregulated genes showed consistent trends with the inflammation stage utilizing an independent CHB dataset. Three (PXN-AS1, RAD9A, SLC17A9) of 27 downregulated genes were found significantly decreased in plasma exosome of CHB patients. ROC analysis revealed that PXN-AS1, RAD9A and SLC17A9 exhibited moderate diagnostic performance in distinguishing CHB from healthy controls, with AUC values of 0.743, 0.762, and 0.665 respectively. A machine learning model, Adaboost classifier, was constructed to detect CHB by combining exosomal expression of PXN-AS1, RAD9A and SLC17A9. The AUC of the model was 0.983 and 0.924 for CHB detection in train and test dataset respectively.

Based on multiple RNA-seq data of tissues and plasma exosomes, we identified PXN-AS1, RAD9A, SLC17A9 as diagnostic biomarkers for CHB detection. The model based on three biomarkers showed potential diagnostic value for detecting CHB. Additional validation with a larger sample size is essential to thoroughly assess the reliability of these three biomarkers and the model's performance.