Predicting the influence of homologous recombination repair deficiency genes on glioma heterogeneity and patient prognosis using multi-omics analysis and machine learning

Background: Glioma is the most common malignant tumor of the central nervous system, and homologous recombination deficiency (HRD) may play a crucial role in its progression. Our study aimed to predict the impact of HRD on glioma heterogeneity and patient prognosis from a multi-omics perspective. Methods: We integrated HRD-related gene expression levels and survival information from The Cancer Genome Atlas (TCGA) and Chinese Glioma Genome Atlas (CGGA) databases. Using a combination of machine learning algorithms, we identified the optimal algorithm and constructed the HRD Index model. After validating the model’s accuracy, we assessed the expression heterogeneity of HRD-related genes in vitro using quantitative polymerase chain reaction (qPCR). Multiple omics analyses, including enrichment analysis, genomics, prediction of immune cell subtype infiltration, and drug sensitivity, were employed to demonstrate the heterogeneity and clinical predictive significance of the HRD Index in glioma. Results: Through algorithm selection, the LASSO-RSF (Least Absolute Shrinkage and Selection Operator – Random Survival Forest) algorithm identified 7 genes (POLR2F, FANCB, PTEN, PLK3, INO80D, PRMT6, and UNG) to construct the HRD Index. Model validation demonstrated excellent accuracy. qPCR results revealed differential expression of these HRD Index genes among different cell lines. Samples grouped by HRD Index showed potential differences in certain cytokine and receptor pathways, as well as varying gene mutation frequencies between groups. Drug sensitivity analysis indicated that the HRD Index could predict treatment efficacy for specific drugs. Conclusion: Our HRD Index model based on these seven genes significantly correlated with clinical prognosis in glioma patients and holds promise for guiding clinical management.