Record 2021 Kidney Transplants Spark Machine Learning Breakthrough in Graft Survival
The United States witnessed an astonishing milestone in 2021, with a record 25,487 kidney transplants performed, according to the latest data report from the Organ Procurement and Transplantation Network and Scientific Registry of Transplant Recipients. This remarkable achievement in kidney transplantation brings hope to thousands of patients in need of life-saving organ transplants.
However, the success of kidney transplants goes beyond the mere number of procedures performed. Five years after transplantation, the survival of transplanted kidneys, known as graft survival, plays a critical role in determining the long-term outcome for recipients. Recent data reveals that among patients aged 18 to 34, graft survival for kidneys from deceased donors stood at 81%, while for individuals above the age of 65, the rate was 68%.
In an effort to improve kidney matching and reduce the risk of graft failure, Malek Kamoun from the Perelman School of Medicine and Ryan Urbanowicz from Cedars-Sinai Medical Center have embarked on a groundbreaking research project employing machine learning strategies. Kamoun and Urbanowicz, along with a team of talented undergraduates from the University of Pennsylvania, have been working diligently to push the boundaries of modern medicine.
Three bright students, Sphia Sadek, Antonios Kriezis, and Aryan Roy, spent their summer contributing to this project remotely. Sadek and Roy were part of the esteemed Penn Undergraduate Research Mentoring Program, where they received a $5,000 award for their invaluable 10-week contributions. Their involvement was made possible through the support of the Center for Undergraduate Research and Fellowships.
Sadek, a third-year student pursuing computer science and cognitive science, joined the project with a strong interest in artificial intelligence and machine learning. Her main focus has been determining the optimal threshold for amino acid mismatches, which are crucial in assessing the risk of graft failure. By stratifying the two groups into low-risk and high-risk categories based on these mismatches, Sadek aims to refine the understanding of graft survival and its implications for different populations, potentially enhancing risk stratification by race and ethnicity.
Leveraging her skills in Python programming and utilizing data from the Scientific Registry of Transplant Recipients (SRTR), Sadek has made significant progress. Her method, under the guidance of Urbanowicz, shows great promise, fueling optimism for improved outcomes in kidney transplantation.
While Sadek delved into the technical aspects of the project, Kriezis and Roy took on different dimensions of the research. As second-year students, their contributions were both vital and impactful. Kriezis, studying systems science and engineering, focused on testing the Akaike Information Criteria (AIC) model, a statistical approach that addresses covariates, such as age, sex, and race. By refining this model, Kriezis aims to improve the accuracy of risk assessments in kidney transplantation.
On the other hand, Roy explored a complementary technique proposed by statistician Keith McCullough. Roy’s approach involved adjusting for residuals in the bins established by the FIBERS algorithm, developed by Kamoun and Urbanowicz. Both Kriezis and Roy’s efforts have provided valuable insights and potential avenues for further research.
The FIBERS algorithm has emerged as a pivotal tool in this groundbreaking work. By automatically identifying bins of amino acid-level mismatches, this evolutionary algorithm effectively categorizes donor-recipient pairs into low-risk and high-risk groups for graft survival. Comparing the outcomes obtained from FIBERS to traditional methods, Kamoun and Urbanowicz observed that FIBERS categorized more patients as low risk, shedding light on previously overlooked amino acid variations that can significantly impact graft failure.
Their research, published recently in the Journal of Biomedical Informatics, presents a compelling case for the integration of machine learning strategies in kidney transplantation. With their innovative approach, Kamoun, Urbanowicz, and their team are paving the way for more refined risk assessments and potential improvements in kidney allocation policies. Their work holds tremendous promise for identifying transplant recipients at higher risk of graft failure, ultimately ensuring better outcomes for patients.
The dedication and hard work of these students, mentored by Kamoun and Urbanowicz, have not only deepened their understanding of machine learning and artificial intelligence but have also provided them with invaluable research experience. Through their participation in the Penn Undergraduate Research Mentoring Program, they have honed their critical thinking, problem-solving, and communication skills, setting themselves up for future success in their academic and professional endeavors.
The pursuit of advancements in kidney transplantation is far from over. Thanks to the tireless efforts of researchers like Kamoun, Urbanowicz, and their team, we can envision a future where the success and survival rates of kidney transplants continue to rise. By harnessing the power of machine learning and pushing the boundaries of medical knowledge, they are providing hope to countless individuals eagerly awaiting life-changing organ transplants.
