Multi-omics analysis identifies RARRES1 as a potential biomarker linked to immunosuppressive microenvironment and its...

Retinoic acid receptor responder 1 (RARRES1) is aberrantly expressed across multiple cancers, its prognostic and immune associations in ovarian cancer (OV) have been increasingly reported, but an integrated multi-omics interpretation across cellular heterogeneity, mutational landscape, and non-invasive radiomics prediction remains insufficiently established. This study aimed to systematically investigate the multi-omics role of RARRES1 in OV and to develop a noninvasive radiomics approach for predicting its expression.

Bulk RNA sequencing (RNA-seq) data from The Cancer Genome Atlas Ovarian Cancer (TCGA-OV) cohort were used to assess associations between RARRES1, survival, and tumor microenvironment (TME) features. Single-cell RNA sequencing dissected cellular heterogeneity and macrophage subpopulations, and weighted gene co-expression network analysis identified RARRES1-associated gene modules. Whole-exome sequencing profiled tumor mutation burden and landscapes. RARRES1 and key immune markers were validated by quantitative reverse transcription polymerase chain reaction (qRT-PCR)/Western blot. A radiomics-based random forest model was constructed to non-invasively predict RARRES1 expression.

High RARRES1 expression was associated with poor overall survival, reduced stromal and immune scores, and increased tumor purity. Functional analyses showed downregulation of antigen presentation, chemokine, interferon, and natural killer (NK) cytotoxicity pathways. Single-cell analysis identified macrophage subpopulations, with M2-like tumor-associated macrophages central and enriched in immunosuppressive and inflammatory pathways. High RARRES1 expression exhibited distinct mutational patterns and weighted gene co-expression network analysis (WGCNA) modules correlated with M1 macrophages and immune activation. Compared with the RARRES1-low group, RARRES1-high tumors showed decreased CD8 and increased CD68 and CD163 expression. The radiomics-based random forest model demonstrated good discrimination and calibration in predicting RARRES1 expression, establishing a direct link between CT-derived imaging features and molecular expression states.

This study establishes a radiogenomic framework for non-invasive prediction of RARRES1 expression in OV. While transcriptomic and immunological findings are consistent with existing evidence, the radiomics-based prediction model offers a clinically applicable strategy to infer tumor immune-molecular states from routine imaging.