Menu
close
Menu
close
Menu
close
Menu
close
Menu
close
Menu
close

Try UroPredict

Click Here

UroPredict

Leverage an innovative machine learning model to predict kidney cancer upstaging and recurrence

Developing the UroPredict algorithms with UroCCR

The goal of UroCCR, the French Kidney Cancer Research Network (Réseau Français de Recherche sur le Cancer du Rein), is to connect a National, multidisciplinary network of medical and scientific professionals who focus on therapeutic management and applied research into kidney cancer.

One of the world’s largest collaborative kidney cancer databases

47

Multidisciplinary clinical teams

3

Countries – France, Belgium, French Guiana

>120

Publications

All clinical, biological, and radiological data from newly diagnosed kidney cancer patients in the UroCCR network are collected in a National, multidisciplinary, clinical and biological database which is shared and used to power collaborative research projects. It is labelled by the French national Cancer Institute (INCa) and referenced by the High Authority of Health (HAS) as a valuable source of real-world data. 

Try the UroPredict calculators

Use predictors with your own data
For research use only – Not for use in diagnostic procedures.

  • Our work with UroCCR over the last three years has significantly advanced its research of RCC and shown the power of AI to provide insights from real-world multimodal data. We are extremely pleased with the findings from this study and look forward to continued collaboration with UroCCR.

    Thierry Colin
    VP, Multimodal Research and Development, SOPHiA GENETICS

  • At UroCCR, our research is centered on the idea that a shared database helps facilitate and expand the use of precision medicine for patients facing kidney cancer. Our work with SOPHiA GENETICS and the results of our latest study demonstrated the benefits of using AI to analyze real-world multimodal data. SOPHiA GENETICS’ technology and network have been key to the success of our research and will be imperative as we continue progressing with increased data inputs.

    Prof. Jean-Christophe Bernhard
    UroCCR Coordinator, Bordeaux University Hospital, France

  • Our work with UroCCR over the last three years has significantly advanced its research of RCC and shown the power of AI to provide insights from real-world multimodal data. We are extremely pleased with the findings from this study and look forward to continued collaboration with UroCCR.

    Thierry Colin
    VP, Multimodal Research and Development, SOPHiA GENETICS

  • At UroCCR, our research is centered on the idea that a shared database helps facilitate and expand the use of precision medicine for patients facing kidney cancer. Our work with SOPHiA GENETICS and the results of our latest study demonstrated the benefits of using AI to analyze real-world multimodal data. SOPHiA GENETICS’ technology and network have been key to the success of our research and will be imperative as we continue progressing with increased data inputs.

    Prof. Jean-Christophe Bernhard
    UroCCR Coordinator, Bordeaux University Hospital, France

Harnessing machine learning to advance kidney cancer research

UroCCR and SOPHiA GENETICS joined forces to develop machine learning models: First to predict prior to surgery whether kidney cancer would upstage, and second to predict whether kidney cancer would recur after surgery.
Read About the Collaboration

Research objectives

To develop a machine learning-based, contemporary, clinically relevant model for pre-operative prediction of renal cell carcinoma pT3a upstaging in patients undergoing nephrectomy for cT1/cT2a renal cell carcinoma.

In simple terms: Develop a model that can predict whether kidney cancer will progress from a localized tumor to a locally advanced tumor (associated with worse prognosis) before the patient undergoes surgery.

To develop a machine learning-based, contemporary, clinically relevant model for prediction of disease-free survival in patients undergoing surgery for localized or locally advanced renal cell carcinoma.

In simple terms: Develop a model that can predict whether a patient’s kidney cancer will recur after they undergo surgery.

Prediction of pT3a upstaging in localized renal cell carcinoma

UroCCR-15
  • Study type: Observational
  • Intervention: Retrospective analysis
  • Clinical Trials Identifier: NCT03293563
  • Publication: Boulenger de Hauteclocque et al. BJU Int. 2023
  • Cohort: 4395 patients treated surgically for renal cell carcinoma between 2000-2019, either by laparoscopic (pure or robot-assisted) partial nephrectomy (PN), open PN, or laparoscopic radical nephrectomy (RN)

Prediction of kidney cancer recurrence after surgery

UroCCR-120
  • Study type: Observational
  • Intervention: Retrospective analysis
  • Clinical Trials Identifier: NCT03293563
  • Publication: Margue et al. NPJ Precis Oncol. 2024
  • Cohort: 3372 patients treated surgically for localized or locally advanced renal cell carcinoma between 2000-2020

Why?

Knowing individuals’ risk of tumor progression/upstaging can guide treatment decisions, enabling selection of the strategy with the best chance of treating the kidney cancer while minimizing side effects and maximizing quality of life.

Research is needed to develop and test such a model.

How?

Machine learning algorithms were designed, trained, and tested to predict kidney cancer upstaging and recurrence after surgery. As some data points might be missing in clinical routine, the algorithms were developed to handle incomplete data.

Read Tech Note on Avoiding Overfitting When Developing ML Models

Model for pT3a upstaging 
(before surgery)

The participants were separated into two cohorts, training (n = 2636) and test (n = 1759). Seven pre-operative features were independently associated with pT3a upstaging and thus included in the predictive model:
  • Tumor size
  • Age
  • Hilar location
  • RENAL score
  • Sex
  • American Society of Anesthesiologists (ASA) score
  • Symptoms at diagnosis
Seven machine learning algorithms were tested and logistic regression was found to be the most effective, with a prAUC of 0.41 on the test dataset and an area under the ROC curve of 0.77.

Model for disease recurrence (after surgery)

The dataset was split into two cohorts, training (n = 2241) and test (n = 1131). 24 clinical, pathological, and biological features were included in the predictive model, including:
  • Tumor size
  • Histological subtype
  • Age
  • Fuhrman grade
  • Sex
  • Pathological primary tumor staging
  • Pathological regional lymph nodes staging
The final machine learning model surpassed the predictive performance of the most commonly used risk scores, with an integrated AUC of 0.79 (95% CI, 0.74–0.83) on the test dataset. The ML model had the further advantage of being able to handle incomplete data.

Applying the pT3a upstaging model

The model predicts individuals’ risk of renal cell carcinoma upstaging before surgery. As an example, the model predicted that one patient had a 32% probability of upstaging, higher than the average of tumor upstaging observed in this study (15%). Using SHAP values, we can interpret this elevated probability with risk factors (ASA score of 2, male sex, tumor size of 8 cm) counterbalanced by protective factors (young age, non-hilar location of tumor).

Applying the disease recurrence model

The model predicts the disease-free survival curve of each patient in the years following surgery, even in presence of incomplete values in predictors. It also assigns each patient to a risk group of recurrence or death within five years after surgery. Finally, using SHAP values, we are able to explain how each multimodal feature contributes to the patient-specific prediction.

What is the impact for patients?

The prediction model has the potential to support decision making around the treatment of clinically localized kidney tumors. Predicting risk of upstaging to pT3a could identify high-risk patients most likely to benefit from preoperative systemic therapy, or even low-risk patients for whom active surveillance could be sufficient.

What is the impact for patients?

The prediction model has the potential to support decision making around the management of patients undergoing localized or locally advanced kidney cancer. Predicting disease-free survival over the years following surgery could enhance the identification of patients candidate for adjuvant therapy or the identification of patients who could benefit from a less intensive post-operative follow-up.

SOPHiA GENETICS products are for Research Use Only and not for use in diagnostic procedures unless specified otherwise.

SOPHiA DDM™ Dx Hereditary Cancer Solution, SOPHiA DDM™ Dx RNAtarget Oncology Solution and SOPHiA DDM™ Dx Homologous Recombination Deficiency Solution are available as CE-IVD products for In Vitro Diagnostic Use in the European Economic Area (EEA), the United Kingdom and Switzerland. SOPHiA DDM™ Dx Myeloid Solution and SOPHiA DDM™ Dx Solid Tumor Solution are available as CE-IVD products for In Vitro Diagnostic Use in the EEA, the United Kingdom, Switzerland, and Israel. Information about products that may or may not be available in different countries and if applicable, may or may not have received approval or market clearance by a governmental regulatory body for different indications for use. Please contact us to obtain the appropriate product information for your country of residence.

All third-party trademarks listed by SOPHiA GENETICS remain the property of their respective owners. Unless specifically identified as such, SOPHiA GENETICS’ use of third-party trademarks does not indicate any relationship, sponsorship, or endorsement between SOPHiA GENETICS and the owners of these trademarks. Any references by SOPHiA GENETICS to third-party trademarks is to identify the corresponding third-party goods and/or services and shall be considered nominative fair use under the trademark law.