Improving the success of kidney transplants with AI-supported precision medicine
For patients with end-stage kidney disease, an organ transplant can be their best hope for recovery. Yet despite medical advances such as immunosuppressive drugs, around 10 % of kidney transplants are rejected within the first year, says Maud Racapé from the French National Institute of Health and Medical Research(opens in new window). “With so many factors involved, predicting rejection is challenging,” says Racapé, coordinator of the EU-TRAIN(opens in new window) project. “Our EU-TRACER tool uses relevant clinical, immune system, genetic and biomarker data to quantify individual risks.” Accessible through a secure web-based interface, EU-TRACER helps clinicians assess patients in real time for early signs of rejection, alongside high or low risk of complications, to avoid unnecessary invasive procedures. “Incorporating non-invasive biomarkers and gene expression data into care parameters could improve prediction accuracy by around 30 %,” adds Racapé.
Multidimensional data for risk stratification
EU-TRAIN combined an existing database of over 5 000 kidney transplant recipients with new data gleaned from two patient studies. In the first, machine-learning techniques and advanced statistical modelling analysed thousands of data points from 554 patients to ascertain if both new and previously identified biomarkers could help predict transplant rejection. “Many of the biomarkers tested, informed by previously published results(opens in new window), did not significantly improve rejection predictions compared to standard monitoring, highlighting the importance of real-world testing,” notes Racapé. “The exception was CD4, a protein that activates other immune cells, alongside so-called ‘circulating anti-HLA DSA’, blood antibodies that attack transplanted organ proteins.”
Leveraging AI-enabled precision medicine
These results informed the second study testing the AI-enabled EU-TRACER tool to evaluate if biomarker-guided monitoring safely reduced the number of biopsies performed within the first year of transplantation. The tool’s algorithm, informed by select clinical, biological and immunological parameters relevant to rejection prediction – including a new promising non-invasive biomarker called donor-derived cell-free DNA (dd-cfDNA) – was applied to 342 new kidney transplant patients. “Instead of doctors having to review separate patient charts, the algorithm aggregates the most predictive parameters,” explains Racapé. Compared with a control group with standard compatibility monitoring, 64 % of biopsies were cancelled in the EU-TRACER group. There were similar rates of rejection, renal function, donor organ loss and death between both groups. “These results show that the EU-TRACER tool can be safely used to avoid invasive protocol biopsies,” notes Racapé.
The transition from research to clinical use
With chronic kidney disease affecting over 100 million Europeans and exerting annual health costs of EUR 140 billion(opens in new window), EU-TRAIN helps address the urgent need for more accurate diagnostics and effective treatments. More broadly, the project contributes to European initiatives to advance personalised medicine(opens in new window) for improved healthcare outcomes. The team is now extending the EU-TRACER Impact Study to assess if the tool can also avoid biopsies for suspected organ rejection. Meanwhile EU-TRACER will be further validated in larger more diverse patient populations, with new biomarkers fed into its predictive models. “Our vision is to expand the platform to other organ transplants – already under way in heart transplantation – to help establish data-driven precision medicine as standard in transplant care,” concludes Racapé.
