Integrated bulk and single-cell RNA sequencing identifies oxidative stress signatures of radiation-induced lung injury in mice through machine learning.

BACKGROUND

Radiation induced lung injury (RILI) is a common complication in patients undergoing thoracic radiotherapy. At present, there are no effective early diagnostic biomarkers, and clinical treatment methods are very limited, which poses a huge challenge to the management of cancer patients. Oxidative stress has been recognized as a key mediator of aging and disease. Therefore, this study integrated multiple omics data in mice and advanced bioinformatics and machine learning methods to systematically analyze the molecular features associated with oxidative stress, and screened for clinically relevant biomarkers and molecular mechanisms of RILI. This study aims to provide a timely and practical theoretical basis for the early diagnosis and targeted intervention of RILI.

METHOD

We implemented a comprehensive approach that integrated both bulk RNA and single-cell RNA sequencing analyses, utilizing advanced bioinformatics methodologies. These encompassed techniques aimed at eliminating batch effects to facilitate smooth data integration, executing differential expression analyses, and applying weighted gene co-expression network analysis (WGCNA). Furthermore, we developed a diagnostic model for RILI utilizing random forest and support vector machine (SVM) algorithms. We also conducted Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Gene Set Enrichment Analysis (GSEA). To evaluate immune cell infiltration, we employed Single-Sample Gene-Set Enrichment Analysis (ssGSEA) alongside the CIBERSORT algorithm. We then investigated the expression and interactions of module genes across various cell populations utilizing data derived from single-cell RNA sequencing. Ultimately, the expression of module genes in irradiated lung tissues were validate by reverse transcription-polymerase chain reaction (RT-PCR) and immunohistochemistry (IHC).

RESULTS

Our study identified a total of 286 differentially expressed genes (DEGs). Among these, we confirmed 61 genes related to oxidative stress (OSRDEGs). We constructed nine co-expression modules, four of which showed a significant association with RILI, encompassing 53 genes from these modules. A diagnostic model with AUC over 0.9 was constructed and further refined to include five key genes: Stk4, Aaas, Ets1, Sesn2, and Kit, which were validated for accuracy through LASSO regression. The model genes were found to be enriched in crucial pathways, particularly the MAPK signaling pathway. A direct relationship between Ets1 and Kit was found, which extended to 20 functionally similar proteins identified through GeneMANIA. Additionally, we noted significant changes in the infiltration patterns of 13 immune cell types, including Activated B cells and Activated CD4 T cells. Sens2 and Kit were found highly expressed in granulocytes and endothelial cells, respectively. In mouse models of RILI, Sesn2 and Aaas were significantly upregulated, whereas Stk4, Ets1, and Kit were downregulated.

CONCLUSION

Our thorough bioinformatics analysis reveals significant molecular events in RILI, identifying 5 key genes and their related signaling pathways. These insights deepen our understanding of the mechanisms underlying the development and progression of RILI and suggest a practical and effective approach for treatment and early diagnosis.