Clear cell renal cell carcinoma (ccRCC) is the most common subtype of kidney cancer, accounting for approximately 70 to 80 percent of all cases. Despite advances in targeted therapy and immunotherapy, prognosis for advanced ccRCC remains poor, and reliable biomarkers for risk stratification are lacking.
Ribosome biogenesis (Ribosis) refers to the complex process of assembling new ribosomes, the cellular machinery responsible for protein synthesis. Ribosome biogenesis is upregulated in many cancers as rapidly dividing cells require high protein synthesis capacity, and dysregulation of this process has been linked to tumor aggressiveness and treatment resistance.
The relationship between ribosome biogenesis and the tumor microenvironment is bidirectional: cancer cells with elevated ribosome production can outcompete immune cells for metabolic resources, potentially suppressing anti-tumor immune responses. This connection makes ribosome biogenesis genes potentially informative for predicting both prognosis and immunotherapy response.
Prior prognostic models in ccRCC have largely focused on single-omics data and individual machine learning algorithms. A more robust approach requires integration of multiple genomic data layers and a consensus framework that identifies features validated across multiple statistical methods, reducing the risk of algorithm-specific overfitting.
The study integrated three layers of genomic data: bulk RNA sequencing, single-cell RNA sequencing (scRNA-seq), and spatial transcriptomics. Bulk RNA-seq from The Cancer Genome Atlas (TCGA) and other cohorts provided population-level gene expression across hundreds of ccRCC samples. scRNA-seq enabled cell-type-specific expression profiling, and spatial transcriptomics mapped gene expression to tissue locations.
Single-cell data allowed identification of which specific cell types within the tumor microenvironment express ribosome biogenesis genes at elevated levels. This resolved whether Ribosis signatures in bulk data reflect expression in tumor cells themselves, immune infiltrates, stromal cells, or a combination of all three.
Spatial transcriptomics data added a geographic dimension, revealing how ribosome biogenesis activity is distributed across the tumor tissue. This enabled identification of spatially distinct domains characterized by high Ribosis activity and assessment of whether these domains co-localize with specific immune cell populations or necrotic regions.
Ribosome biogenesis-related genes were curated from the MSigDB gene set database and literature-derived gene sets. The intersection of genes differentially expressed in ccRCC with established Ribosis pathway members formed the candidate feature set for subsequent prognostic modeling.
A critical innovation in this study was the use of 118 combinations of machine learning algorithms to identify the most stable and generalizable prognostic signature. This approach systematically evaluated combinations spanning penalized regression methods such as LASSO and ridge regression, tree-based ensembles such as random forest and gradient boosting, support vector machines, and survival-specific methods including Cox regression and survival random forest.
Each of the 118 algorithm combinations was trained on a discovery cohort and evaluated on multiple independent validation cohorts. The concordance index (C-index), a measure of model discrimination equivalent to AUC for survival data, was used to rank algorithm performance. The combination achieving the highest average C-index across all validation cohorts was selected.
This consensus approach guards against the common problem of selecting a signature that performs well for one specific algorithm or cohort but fails to generalize. By requiring consistent performance across 118 alternatives and multiple external datasets, the final signature is more likely to reflect genuine biological signal rather than statistical artifact.
The output of the final selected algorithm was a Ribosome Biogenesis Risk Score (RBRS), computed as a weighted linear combination of the selected gene expression values. Each patient's RBRS was used to stratify them into high-risk and low-risk groups for overall survival analysis.
The final RBRS model achieved a C-index of 0.68 in the primary TCGA validation cohort and maintained similar performance across multiple independent external validation datasets. This level of discrimination is meaningful in the context of survival prediction, where c-indices above 0.65 are considered clinically informative for complex diseases like cancer.
High RBRS was significantly associated with worse overall survival across all validation cohorts in multivariable Cox regression analyses that adjusted for clinical covariates including tumor stage, grade, and patient age. This confirms that the RBRS adds prognostic information beyond what standard clinical parameters already capture.
RBRS high patients showed enrichment in immune-suppressed tumor microenvironment signatures, including reduced T cell infiltration, lower cytotoxic activity scores, and elevated expression of immune checkpoint molecules such as PD-L1. This suggests that high ribosome biogenesis activity may contribute to an immune-cold microenvironment unfavorable for immunotherapy response.
Single-cell and spatial transcriptomics analyses confirmed that the ribosome biogenesis signature was predominantly expressed in malignant epithelial cells rather than immune or stromal cells, with spatial clustering of high-Ribosis regions coinciding with areas of high tumor cell density and low immune infiltration.
The association between high RBRS and an immune-suppressed microenvironment has direct implications for treatment stratification. Patients with high RBRS may derive less benefit from immune checkpoint inhibitors, which depend on sufficient pre-existing immune infiltration and T cell activity to be effective.
Conversely, patients with low RBRS and a more immune-active tumor microenvironment may be prioritized for immunotherapy-based regimens. This hypothesis aligns with observations in clinical trials where baseline tumor immune infiltration correlated with immunotherapy response in ccRCC and other cancers.
The connection between ribosome biogenesis and tumor aggressiveness also suggests that ribosome synthesis inhibitors, currently in preclinical development, could have therapeutic relevance in high-RBRS ccRCC. Targeting ribosome assembly could simultaneously suppress tumor cell proliferation and potentially reverse immune suppression.
A limitation of the multi-algorithm consensus approach is that it requires substantial computational resources and a large number of well-annotated training and validation cohorts. The method is likely to be most impactful for diseases where multiple large public datasets are available, as is the case with ccRCC, and may be harder to apply to rarer cancer types.
This study demonstrates that integrating bulk RNA-seq, scRNA-seq, and spatial transcriptomics with a 118-algorithm machine learning consensus framework can identify a robust and biologically interpretable prognostic signature in ccRCC based on ribosome biogenesis genes.
The RBRS provides independent prognostic value beyond clinical staging, characterizes the immune microenvironment, and may help predict differential response to immunotherapy versus targeted therapy, moving toward a precision oncology framework for ccRCC treatment decisions.
The multi-omics and multi-algorithm methodology developed here is generalizable to other cancer types and other biological processes beyond ribosome biogenesis. It provides a template for developing reproducible and externally validated molecular signatures in the era of large-scale genomic data availability.
Future validation in prospective cohorts linked to clinical trial outcomes is needed to establish the RBRS as a clinically actionable biomarker. Integration with existing clinical risk scores such as the International Metastatic RCC Database Consortium criteria could further refine risk stratification in advanced ccRCC.