CORDIS
EU research results

CORDIS

English EN

News

A novel approach of machine learning identifies patients at risk of losing kidney transplants due to a common pathology

Transplant glomerulopathy is a common pathology in kidney transplantation that is associated with graft loss. French researchers identified 5 groups of patients with distinct characteristics - including different renal transplant survival chances. A machine learning approach was applied to a database to get these results. This promising approach is being translated into a clinician-accessible tool to assess the risk of graft loss associated with transplant glomerulopathy.

Scientific advances

© courtesy of Labs Explorer

New research could help clinicians determine which patients are at high risk of kidney transplant failure when presenting glomerulophathy, a disease that usually occurs after a kidney transplant. The results are published in March in the prestigious Journal of the American Society of Nephrology (JASN). A disease known as transplant glomerulopathy was first described in kidneys transplanted 50 years ago, but its specificities have not yet been defined. The disease is common and unfortunately associated with an unfavourable prognosis. It affects the functional units (i.e. glomeruli) of the transplanted kidney. Transplant glomerulopathy affects nearly half of all kidney transplants. Some of its mechanisms that can be identified through a simple microscope may also be due to other pathological processes, making its diagnosis difficult. To provide clarification, Dr. Olivier Aubert, Prof. Alexandre Loupy and their colleagues at the Paris Translational Research Center for Organ Transplantation analyzed complete results, clinical and immunological data and observed patient survival rates. As a result, they were able to identify and characterize distinct groups of patients with transplant glomerulopathy. Dr. Loupy said: "Not all patients with the same disease are the same, and machine learning approaches can help us identify different risk profiles for graft loss in a heterogeneous patient population." By applying a machine learning approach to the data of 385 patients diagnosed through biopsies, the researchers identified five groups of patients with distinct characteristics. In particular, different results in terms of survival rates of transplanted kidneys. "Our research shows how we could interrogate a dataset with multiple levels of complexity in patients with the same disease, in order to distinguish different profiles with different survival probabilities of transplanted organs," said Dr. Aubert. The team has also developed a tool that can be easily used by clinicians to enter patients' clinical, histological and immunological results and to obtain a prediction of their clinical results (https://dyshinyapps.shinyapps.io/Archetype_Shiny/). The tool would allow patients' treatments to be individualized according to their risk of transplant failure.

Keywords

Machine Learning, artificial intelligence, transplantation, kidney, glomerulopathy, graft, Alexandre Loupy

Countries

France, United States