Aggressiveness in clear cell renal cell carcinoma highlighted through CBX4 radiogenomic evidence: a cross-sectional study

Introduction

Radiogenomics is an emerging field in radiology that links imaging characteristics to the genetic profile of diseases (1,2). Imaging phenotypes reflect molecular processes observable through imaging techniques (1,2). The Cancer Genome Atlas (TCGA) Research Network provides extensive data on gene expressions and mutations associated with clear cell renal cell carcinoma (ccRCC) (3,4).

Radiogenomic analysis offers several advantages: by examining the entire tumor, it provides data that may not always be accessible through genomic testing on biopsy samples. This approach captures information about the tumor’s overall heterogeneity, which is crucial for disease prognosis. Genomic tests performed on small biopsy samples may fail to capture significant genomic data relevant to prognosis, potentially leading to non-personalized therapeutic strategies (5). Additionally, it has been demonstrated that tumor cells with similar genotypes can exhibit different phenotypes, a challenge that radiogenomics could help address (6). Radiogenomic correlations have also been explored in the context of body composition, quantified using a computed tomography (CT)-based approach, and genomic data in ccRCC patients (7).

Polycomb group (PcG) proteins are key transcriptional repressors that epigenetically modify chromatin, playing an essential function in regulating cell differentiation, senescence, and survival (8). Polycomb chromobox (CBX) proteins are part of the PcG family and consist of five members (CBX2, CBX4, CBX6, CBX7, and CBX8). Depending on the cellular environment, CBX proteins operate through the polycomb repressive complex 1 (PRC1) to mediate either tumor-suppressing or tumor-promoting effects (9). It has been demonstrated that CBX4 plays an oncogenic role in ccRCC. CBX4 expression was found to be elevated and associated with poor prognosis. CBX4 facilitated cell proliferation and migration by interacting with Histone Deacetylase 1 (HDAC1) to transcriptionally suppress the expression of the tumor suppressor Kruppel-like factor 6 (KLF6) (10). Furthermore, inhibition of CBX4 enhanced cell apoptosis induced by treatment with the HDAC inhibitor trichostatin A (TSA) (10). These findings highlight CBX4 as a promising prognostic marker and a potential therapeutic target in ccRCC.

Radiogenomics has rapidly expanded in recent years, linking imaging-derived phenotypes to gene expression programs with potential implications for individualized risk prediction and treatment selection in ccRCC. Previous radiogenomic studies in ccRCC have mainly investigated imaging-genomic correlations involving genes such as prolyl 4-hydroxylase subunit alpha 3 (P4HA3) , members of the GTPase domain of the immune associated nucleotide binding protein (GIMAP) family, and disintegrin and metalloproteinase domain-containing protein (ADAM12) (11-13). These studies demonstrated that radiological features including tumor size, necrosis, and margin definition reflect different molecular pathways and prognostic behaviors. However, despite CBX4’s known oncogenic potential and association with poor prognosis, no prior research has explored its radiogenomic profile. Addressing this gap provides novel insights into how CBX4 expression may relate to imaging features indicative of tumor aggressiveness. Although multiple genes have been implicated, we specifically focused on CBX4 given its documented oncogenic potential in ccRCC and the lack of prior radiogenomic characterization. Given CBX4’s established role in promoting oncogenic behaviors, such as enhanced proliferation and migration, understanding its correlation with imaging features could provide valuable insights into disease prognosis and management. We present this article in accordance with the STROBE reporting checklist (available at https://tro.amegroups.com/article/view/10.21037/tro-25-10/rc).

Methods

This was a retrospective cross-sectional clinical research study. Given the exploratory radiogenomic correlation objective and the available sample size (n=206), the cohort is adequate for hypothesis-generating analysis. Data collection included clinical variables, CT phenotypes and CBX4 RNA expression extracted from TCGA. All clinical, imaging, and RNA-sequencing data were obtained from The Cancer Imaging Archive (TCIA) and TCGA repositories. TCGA-Kidney Renal Clear Cell Carcinoma (KIRC) dataset includes imaging and clinical data collected approximately between 2006 and 2013. Data collection and download were performed on November 1, 2019, and data collection continued for the following five months. Patients were selected from the TCGA-KIRC project based on predefined inclusion and exclusion criteria. Inclusion criteria were histologically confirmed ccRCC, availability of pre-treatment contrast-enhanced CT images, and corresponding RNA-sequencing data in the TCGA-KIRC collection. Exclusion criteria included missing or poor-quality imaging data or absence of matched gene expression data. The final study cohort consisted of 206 patients who met all eligibility criteria (Figure 1). For variables with missing data, analyses were performed using available cases only (complete-case analysis). No imputation of missing values was performed. CBX4 expression data were retrieved from RNA-sequencing data in the TCGA-KIRC project. Expression values were log2-transformed normalized counts. Patients were divided into CBX4-positive and CBX4-negative groups using the median CBX4 expression value as a threshold (all cases ≥ median classified as positive).

Figure 1 Flowchart of the patient selection process. A total of 267 patients were initially identified from the TCGA-KIRC database. Patients with missing or poor-quality CT imaging or without matched gene expression data were excluded. After applying the exclusion criteria, 206 patients were included in the final analysis. CT, computed tomography; TCGA-KIRC, The Cancer Genome Atlas-Kidney Renal Clear Cell Carcinoma.

CT images and data of ccRCC patients were retrieved from TCIA. TCIA project received approval of the Institutional Review Board. This subsequent retrospective analysis was performed on publicly available, anonymized data and did not require further review due to the protections previously implemented by TCIA.

TCGA

TCGA is a comprehensive resource detailing the genetic alterations across more than 20 types of cancer, including ccRCC. This initiative was funded by the National Cancer Institute and the National Human Genome Research Institute (NHGRI). Collaborating institutions contributed numerous tissue specimens, which were subjected to extensive genomic analysis using multiple platforms.

TCIA is a repository of anonymized pretreatment medical images stored in DICOM format. Supported by the National Cancer Institute, this archive links medical images and tissue samples through a unique identifier corresponding to TCGA tissue specimens. The data is publicly available for download (14).

Imaging features

The CT characteristics assessed for each ccRCC lesion included: tumor size (mm), margin definition (categorized as well-defined or ill-defined), composition (classified as solid or cystic), and necrosis (evaluated in the solid portion of the tumors with percentages of 0%, 1–33%, 34–66%, or >66%) (7,11-13). Additional parameters included growth pattern (endophytic, <50% exophytic, or ≥50% exophytic), calcification (present or absent), laterality (left or right kidney), and the presence of collateral vascular supply (enlarged renal capsular veins visible on CT) (7,11-13). Other important features assessed were infiltration of adjacent tissues, invasion of the collecting system, intralesional hemorrhage, hydronephrosis, renal artery thrombosis, and renal vein thrombosis (7,11-13). Two further CT criteria were perirenal fat stranding and Gerota’s fascia thickening, both categorized as absent or present (7,12,13). Tumor size was determined by measuring the maximum axial diameter on postcontrast images (7,11-13). A tumor was considered to have well-defined margins if more than 90% of its circumference appeared sharply outlined in postcontrast images, including its interface with the renal parenchyma, collecting system, sinus, and perinephric fat. The window width and level settings used were W: 400 and L: 50 (7,11-13). Tumors containing ≥50% cystic spaces with fluid attenuation values ≤20 Hounsfield units (HU) were classified as cystic. In contrast, tumors with no cystic component or with less than 50% of the volume occupied by cystic spaces were considered solid (7,11-13). Necrotic regions were identified as hypodense areas lacking contrast enhancement, with no visible walls, and showing indistinct boundaries, thus distinguishing necrosis from cystic components (7,11-13). Tumor necrosis in solid tumors was examined during the nephrographic or excretory phases (7,11-13). Calcifications were identified as high-density plaques or spots. In uncertain cases, calcifications were confirmed if the maximum value exceeded 60 HU (7,11-13). Intralesional hemorrhage was detected by intratumoral areas with HU values consistent with blood (+30 to +80 HU) (7,11-13). To differentiate calcifications from hemorrhagic areas with similar HU values, two experienced radiologists specializing in oncologic imaging (F.G., 9 years of experience; C.A.M., 13 years of experience) evaluated the morphological characteristics of the lesions (7,11-13). Tumor infiltration was defined as cancer extending into adjacent healthy tissue, observed during postcontrast phases (7,11-13). Hydronephrosis was diagnosed when urinary tract dilation was detected in postcontrast phases (7,11-13). Renal artery or vein thrombosis was confirmed by identifying filling defects within the vessel lumen on postcontrast images (7,11-13). Collecting system invasion was determined by detecting filling defects in the system during the excretory phase (7,11-13).

Image analysis procedure

All CT images were reviewed through Horos software (v.4.0.0 RC2). Two board-certified radiologists with 9 and 13 years of experience in oncologic imaging independently evaluated all 206 cases. Both readers were blinded to CBX4 expression status and clinical outcomes. After independent assessment, any discrepancies between the two readers were resolved by consensus in a joint session. To assess inter-observer reliability, Cohen’s kappa coefficient (κ) was calculated for categorical features and intraclass correlation coefficient (ICC, two-way random, absolute agreement) for continuous variables such as tumor size. The overall inter-observer agreement was substantial to excellent (κ =0.78–0.86; ICC =0.91, 95% confidence interval: 0.88–0.94).

Statistical analyses

Descriptive statistics included frequencies and proportions for categorical variables, and medians with interquartile ranges (IQRs) for continuous variables. The relationship between CBX4 expression (positive vs. negative) and clinicopathological characteristics, as well as CT-based tumor features was investigated. Wilcoxon rank sum test, Pearson’s Chi-squared test, and Fisher’s exact test were used to examine the statistical significance of differences in medians and proportions among the patient cohort stratified according to CBX4 expression. All tests were two-sided with a level of significance set at P<0.05. R software environment for statistical computing and graphics (version 4.1.2, R foundation for Statistical Computing, Vienna, Austria) was used for all analyses.

Ethical statement

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This subsequent retrospective analysis was on the publicly available, anonymized data and did not require further review due to previous protections implemented by TCIA.

Results

In this work, we evaluated radiological features of CBX4 expression in ccRCC patients. A significant association was found with primary tumor size (P=0.002), tumor grade (P=0.04), more advanced tumor American Joint Committee on Cancer (AJCC) stage (P=0.02), collateral vascular supply (P=0.03), ill-defined margins (P=0.01), exophytic growth pattern ≥50% (P=0.03), and signs of infiltration (P=0.01). These findings suggest that CBX4 expression correlates with radiological features indicative of increased tumor aggressiveness (Figure 2). Conversely, no significant differences were observed between CBX4-positive and CBX4-negative tumors in other CT features typically associated with aggressive ccRCC behavior, such as tumor necrosis, renal vein or artery thrombosis, collecting system invasion, perirenal fat stranding, Gerota’s fascia thickening, intralesional hemorrhage, and hydronephrosis (all P>0.05; Table 1).

Figure 2 Axial CT image during venous phase showing ccRCC with CBX4 gene expression with signs of infiltration of the transverse colon (light orange arrow in A), ill-defined margins (yellow arrow in A), larger primary tumor size with maximum axial diameter of 12.059 cm (B), exophytic growth pattern ≥50% (green arrow in B), and venous collateral vessels (red arrow in C). CBX4, chromobox 4; ccRCC, clear cell renal cell carcinoma; CT, computed tomography.

Data from an overall cohort of 206 patients were extracted, of whom 87 (42.2%) had a positive CBX4 expression (Table 1).

Among clinical-pathological characteristics, patients with a positive CBX4 expression presented with larger tumors {median primary tumor size: 60 [47–90] vs. 49 [36–75] mm, P=0.002}, and more frequently harbored high-grade (67.8% vs. 53.8%, P=0.04) and locally-advanced or advanced/metastatic (44.8% vs. 35.3%, P=0.02) tumors at final pathology compared to those with a negative CBX4 expression (Table 1).

Among CT-based features evaluated preoperatively, renal tumors in patients with a positive CBX4 expression more frequently showed ill-defined margins (46.0% vs. 29.1%, P=0.01), signs of infiltration (5.8% vs. 0%, P=0.01), an exophytic growth pattern higher than 50% (74.7% vs. 56.8%, P=0.03), and harbored a higher amount of collateral vascular supply (65.9% vs. 50.9%, P=0.03) compared to patients with a negative CBX4 expression (Table 1).

Discussion

All radiogenomic features of CBX4 expression in ccRCC are indicative of tumor aggressiveness. It is well established that larger tumor size in ccRCC is associated with reduced disease-specific survival (DSS) and progression-free survival (PFS) (15). An ill-defined margin and tumor infiltration are established CT features associated with overall aggressive ccRCC, as noted in previous studies (16-18). The presence of collateral vascular supply was significantly associated with high-grade tumors, aggressive clinicopathological parameters, and worse prognosis (19,20). While CBX4-positive tumors more frequently exhibited ≥50% exophytic growth in our cohort, this finding should be interpreted as hypothesis-generating. CBX4’s known role in promoting cell proliferation, migration, and suppression of tumor suppressive transcriptional programs may plausibly favor outward expansion along low-resistance tissue planes; however, a direct mechanistic link between CBX4 activity and exophytic tumor geometry has not yet been demonstrated and requires biological validation. Among other imaging features, tumor necrosis was evaluated and did not significantly differ between CBX4-positive and CBX4-negative tumors (P=0.20). Quantitative vascularity, instead, could not be reliably assessed because TCIA does not provide standardized information on contrast medium volume/protocol, and not all contrast-enhanced phases were consistently available for all patients. Future prospective datasets with full multiphase acquisition and protocol uniformity will be essential to determine whether more granular vascular imaging biomarkers correlate with CBX4 expression.

Primary tumor size and ill-defined margins are radiogenomic features typically associated with ccRCC expressing P4HA3 and ccRCC expressing ADAM12, while infiltration is a characteristic shared by ccRCC expressing P4HA3 and ccRCC expressing GIMAP family genes (11-13). Additionally, a more advanced AJCC tumor stage is a radiogenomic feature typical of P4HA3 expression (11).

However, these genes exhibit a different radiological phenotype compared with ccRCC with CBX4 expression. Specifically, ccRCC with CBX4 expression, compared with ccRCC with P4HA3 expression, shows the presence of collateral vascular supply and an exophytic growth pattern ≥50% (11).

The difference between ccRCC expressing GIMAP family genes and ccRCC expressing CBX4 lies in the fact that ccRCC with CBX4 expression presents primary tumor size, ill-defined margins, venous collateral vascular supply, exophytic growth pattern ≥50%, and more advanced AJCC tumor stage (12).

Meanwhile, the difference between ccRCC expressing ADAM12 and ccRCC expressing CBX4 is that ccRCC with ADAM12 expression shows tumor necrosis and collecting system invasion, whereas ccRCC with CBX4 expression presents collateral vascular supply, exophytic growth pattern ≥50%, infiltration, and more advanced AJCC tumor stage (13).

In radiogenomics, ill-defined tumor margins has also been observed in ccRCC with breast cancer susceptibility gene 1 (BRCA1) associated protein 1 (ubiquitin carboxy-terminal hydrolase) (BAP1) (21) and in ccRCC with high methylation levels of tumor suppressor runt related transcription factor 3 (RUNX3) (22). Specifically, BAP1 mutation, in addition to being associated with ill-defined margins, is linked to renal vein invasion, intratumoral calcifications, and a high Fuhrman grade (21). Conversely, ccRCC with high methylation levels of RUNX3 exhibits ill-defined margins, left-sided tumors, and the presence of intratumoral vascularity (22). CcRCC with CBX4 expression shows renal vein thrombus/infiltration at the threshold of statistical significance (P=0.054) but does not exhibit tumor calcifications as seen in ccRCC with BAP1 mutations, nor does it present left-sided tumors as observed in cases with RUNX3 methylation (21,22).

The interaction between CBX4 and other oncogenic gene programs (e.g., P4HA3, ADAM12, GIMAP family) remains unknown, and whether their radiogenomic phenotypes converge mechanistically or represent independent parallel pathways requires dedicated integrative studies. Additionally, rare aggressive traits such as infiltration occurred in a small number of CBX4-positive tumors (n=5), and therefore these subgroup results should be interpreted with caution due to limited power and require validation in larger multi-institutional cohorts.

This study, while providing valuable insights into the radiogenomic features associated with CBX4 expression in ccRCC, has several limitations that should be acknowledged. The retrospective design inherently introduces potential selection bias, as the cohort was drawn from publicly available datasets, which may not fully represent the broader population of ccRCC patients. Additionally, the reliance on imaging data and clinical annotations from TCIA and TCGA, while robust, limits the ability to validate findings prospectively or in a clinical setting.

The sample size, although significant, may still be insufficient to detect subtle associations or rare features of CBX4 expression. Due to the retrospective design and the use of publicly accessible TCGA/TCIA repositories, the present cohort may not fully represent the broader ccRCC population encountered in routine clinical practice. Patients included in these databases are highly selected, may originate predominantly from academic referral centers. Therefore, the generalizability of these results is limited, and prospective validation using multi-institutional cohorts with more diverse patient demographics and imaging acquisition settings will be essential. Furthermore, a formal inter-observer variability assessment was not performed in this study. Although two radiologists with dedicated oncologic imaging experience evaluated all CT features, the absence of an inter-rater agreement analysis may introduce variability and reduce repeatability. Additionally, TCGA/TCIA datasets lack detailed clinical information (e.g., comorbidities, treatment history, systemic therapy exposure), and therefore potential confounding variables could not be adjusted for. Additionally, the CT datasets were collected from various institutions using non-uniform acquisition parameters, potentially leading to variability in image characteristics and their subsequent evaluation. Prospective studies incorporating standardized image assessment protocols and comprehensive clinical annotation will be necessary to improve methodological robustness and enable appropriate multivariable modeling. Moreover, given the observational retrospective design, this study identifies correlations only and cannot infer causation. Whether CBX4 expression mechanistically drives the observed radiologic aggressiveness, or whether both arise from shared molecular programs or upstream regulatory pathways, cannot be determined from these data. Future mechanistic studies together with prospective validation will be required to elucidate causality and biological directionality.

Finally, while this study establishes correlations between CBX4 expression and certain radiological features, it cannot determine causation or the underlying mechanisms driving these associations. Prospective studies with larger, more diverse cohorts and experimental validation are necessary to confirm these findings and to explore their potential implications for personalized treatment strategies in ccRCC.

Conclusions

In this study, we found that CBX4 expression in ccRCC is significantly associated with imaging features indicative of tumor aggressiveness. CBX4-positive tumors were larger, more frequently high-grade and advanced stage, and more often showed ill-defined margins, ≥50% exophytic growth, collateral vascular supply, and signs of infiltration. These radiogenomic correlations suggest that CBX4 expression may serve as a non-invasive marker of more aggressive disease, supporting its potential role in prognostic stratification. However, prospective studies with larger, more diverse cohorts and experimental validation are required to confirm these findings and evaluate their implications for personalized treatment strategies in ccRCC.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://tro.amegroups.com/article/view/10.21037/tro-25-10/rc

Peer Review File: Available at https://tro.amegroups.com/article/view/10.21037/tro-25-10/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tro.amegroups.com/article/view/10.21037/tro-25-10/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki its subsequent amendments. This subsequent retrospective analysis was on the publicly available, anonymized data and did not require further review due to previous protections implemented by TCIA.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Alessandrino F, Krajewski KM, Shinagare AB. Update on Radiogenomics of Clear Cell Renal Cell Carcinoma. Eur Urol Focus 2016;2:572-3. [Crossref] [PubMed]
2. Pinker K, Shitano F, Sala E, et al. Background, current role, and potential applications of radiogenomics. J Magn Reson Imaging 2018;47:604-20. [Crossref] [PubMed]
3. Comprehensive molecular characterization of clear cell renal cell carcinoma. Nature 2013;499:43-9. [Crossref] [PubMed]
4. Akin O, Elnajjar P, Heller M, et al. The Cancer Genome Atlas Kidney Renal Clear Cell Carcinoma Collection (TCGA-KIRC) (Version 3) [Data set]. The Cancer Imaging Archive. 2016. Available online: https://doi.org/10.7937/K9/TCIA.2016.V6PBVTDR
5. Gerlinger M, Rowan AJ, Horswell S, et al. Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N Engl J Med 2012;366:883-92. [Crossref] [PubMed]
6. Kreso A, O'Brien CA, van Galen P, et al. Variable clonal repopulation dynamics influence chemotherapy response in colorectal cancer. Science 2013;339:543-8. [Crossref] [PubMed]
7. Greco F, Panunzio A, Bernetti C, et al. The Radiogenomic Landscape of Clear Cell Renal Cell Carcinoma: Insights into Lipid Metabolism through Evaluation of ADFP Expression. Diagnostics (Basel) 2024;14:1667. [Crossref] [PubMed]
8. Mozgova I, Hennig L. The polycomb group protein regulatory network. Annu Rev Plant Biol 2015;66:269-96. [Crossref] [PubMed]
9. Ma RG, Zhang Y, Sun TT, et al. Epigenetic regulation by polycomb group complexes: focus on roles of CBX proteins. J Zhejiang Univ Sci B 2014;15:412-28. [Crossref] [PubMed]
10. Jiang N, Niu G, Pan YH, et al. CBX4 transcriptionally suppresses KLF6 via interaction with HDAC1 to exert oncogenic activities in clear cell renal cell carcinoma. EBioMedicine 2020;53:102692. [Crossref] [PubMed]
11. Greco F, Panunzio A, Tafuri A, et al. CT-Based Radiogenomics of P4HA3 Expression in Clear Cell Renal Cell Carcinoma. Acad Radiol 2024;31:902-8. [Crossref] [PubMed]
12. Greco F, Panunzio A, Tafuri A, et al. Radiogenomic Features of GIMAP Family Genes in Clear Cell Renal Cell Carcinoma: An Observational Study on CT Images. Genes (Basel) 2023;14:1832. [Crossref] [PubMed]
13. Greco F, Panunzio A, Bernetti C, et al. Exploring the ADAM12 Expression in Clear Cell Renal Cell Carcinoma: A Radiogenomic Analysis on CT Imaging. Acad Radiol 2024;31:3672-7. [Crossref] [PubMed]
14. CIP TCGA Radiology Initiative—The Cancer Imaging Archive (TCIA) Public Access—Cancer Imaging Archive Wiki [Internet]. 2019. Available online: https://wiki.cancerimagingarchive.net/display/Public/CIP+TCGA+Radiology+Initiative
15. Hötker AM, Karlo CA, Zheng J, et al. Clear Cell Renal Cell Carcinoma: Associations Between CT Features and Patient Survival. AJR Am J Roentgenol 2016;206:1023-30. [Crossref] [PubMed]
16. Peña-Llopis S, Vega-Rubín-de-Celis S, Liao A, et al. BAP1 loss defines a new class of renal cell carcinoma. Nat Genet 2012;44:751-9. [Crossref] [PubMed]
17. Fukatsu A, Tsuzuki T, Sassa N, et al. Growth pattern, an important pathologic prognostic parameter for clear cell renal cell carcinoma. Am J Clin Pathol 2013;140:500-5. [Crossref] [PubMed]
18. Ro JY, Ayala AG, Sella A, et al. Sarcomatoid renal cell carcinoma: clinicopathologic. A study of 42 cases. Cancer 1987;59:516-26. [Crossref] [PubMed]
19. Greco F, Quarta LG, Bernetti C, et al. Composition of perinephric fat and Fuhrman grade in clear cell renal cell carcinoma: the role of peritumoral collateral vessels. Appl Sci 2021;11:3941.
20. Suo X, Chen J, Zhao Y, et al. Clinicopathological and radiological significance of the collateral vessels of renal cell carcinoma on preoperative computed tomography. Sci Rep 2021;11:5187. [Crossref] [PubMed]
21. Shinagare AB, Vikram R, Jaffe C, et al. Radiogenomics of clear cell renal cell carcinoma: preliminary findings of The Cancer Genome Atlas-Renal Cell Carcinoma (TCGA-RCC) Imaging Research Group. Abdom Imaging 2015;40:1684-92. [Crossref] [PubMed]
22. Cen D, Xu L, Zhang S, et al. Renal cell carcinoma: predicting RUNX3 methylation level and its consequences on survival with CT features. Eur Radiol 2019;29:5415-22. [Crossref] [PubMed]