Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


kleinknecht final.pdf (1.24 MB)

ETD Abstract Container

Abstract Header

Machine Learning and Computational Methods for Evaluating Kidney Graft Allocation

Kleinknecht, Justin
http://rave.ohiolink.edu/etdc/view?acc_num=bgsu1589471926006468

Abstract Details

2020, Master of Science (MS), Bowling Green State University, Computer Science.
Kidney transplantation is the most effective long-term solution for renal disease. Unfortunately, there are a multitude of factors that determine how compatible a donor kidney will be with a recipient’s body. Current measures of compatibility involve examination of individual Human Leukocyte Antigens (HLAs) and comparing the number of identical antigens between the donor and recipient. Generally, a lower number of mismatches between antigens results in higher survival time after a graft. An emerging, more insightful, metric for predicting post graft survival time is immunogenicity. In addition, there remain open questions regarding the role of machine learning in the kidney donation and allocation process. Based on these ideas, this thesis pursues two research questions: “What is the role of machine learning in kidney donation and allocation?” and “Can it be computationally demonstrated that it is possible to achieve improved kidney-donor pairings using immunogenicity calculations rather than HLA mismatches?” Investigations into the first question results in the evaluation of logistic regression, random forest, and multi-layer perceptron models in predicting whether or not a particular donor-recipient pair is a “good” match based on the number of antigen mismatches. In this case, either zero or one mismatch is considered as “good”. The random forest and multi-layer perceptron algorithms perform best with balanced accuracy scores above 90% and F1 scores of nearly 90%. Feature selection is also applied within this process, resulting in insights regarding the importance of various variables and techniques as it relates to kidney donation. The second question leads to the development of a preliminary simulation model that systematically separates donor-recipient pairs and the recombines in order to achieve superior outcomes. In this case, the measure evaluated includes two immunogenicity calculations named Electrostatic Mismatch Score (EMS) and Hydrophobic Mismatch Score (HMS), as these measures have been shown to potentially indicate improved outcomes even when there are a large number of HLA mismatches between donors and recipients. Through various computational improvements, initial trials demonstrate that it is possible to rematch existing donors and recipients to achieve reduced HMS and EMS scores for roughly 70%–90% of donor-recipient pairs.
Robert Green, Ph.D. (Advisor)
Dulat Bekbolsynov, Ph.D. (Committee Member)
Michael Decker, Ph.D. (Committee Member)
Robert Dyer, Ph.D. (Committee Member)
68 p.
Computer Science
Machine Learning; Kidney Allocation; Simulation

Recommended Citations

Citations

  • Kleinknecht, J. (2020). Machine Learning and Computational Methods for Evaluating Kidney Graft Allocation [Master's thesis, Bowling Green State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=bgsu1589471926006468

    APA Style (7th edition)

  • Kleinknecht, Justin. Machine Learning and Computational Methods for Evaluating Kidney Graft Allocation. 2020. Bowling Green State University, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=bgsu1589471926006468.

    MLA Style (8th edition)

  • Kleinknecht, Justin. "Machine Learning and Computational Methods for Evaluating Kidney Graft Allocation." Master's thesis, Bowling Green State University, 2020. http://rave.ohiolink.edu/etdc/view?acc_num=bgsu1589471926006468

    Chicago Manual of Style (17th edition)

bgsu1589471926006468
447
© 2020, all rights reserved.
This open access ETD is published by Bowling Green State University and OhioLINK.