Honors Projects
Explainable AI for Liver Transplant Survival Prediction: Integrating Immunological Mismatch Features
Abstract
Liver Transplantations are crucial treatment for end-stage liver disease. However, a persistent deficit of donor organs necessitates maximizing the utility of each available graft to minimize failure rates. We evaluated whether donor–recipient molecular immunogenicity metrics - Electrostatic and Hydrophobic Mismatch Scores (HMS/EMS) and eplet-based counts - improve post–liver-transplant survival prediction. The analytic cohort comprised adult, first time, single-organ deceased-donor transplants drawn from Scientific Registry of Transplant Recipients; follow-up was truncated at five years, and the endpoint was all-cause graft failure (earliest of graft failure or death; otherwise, censored). HLA variables were derived via high- resolution conversion and molecular mismatch computations extending HLAMatchmaker, providing continuous EMS/HMS features grounded in biophysical differences. We compared Cox proportional hazards (CPH), Random Survival Forests (RSF), and DeepHit models on matched splits. Adding HLA metrics to CPH models produced a minimal C- index change with identical Integrated Brier Score and none of the HLA coefficients were significant. In RSF Models, the HLA-augmented model did not improve discrimination. In DeepHit, HLA features yielded a small mean lift that was within fold variability. Explainability analyses consistently placed clinical covariates at the top, with DR-locus physicochemical metrics registering small, secondary contributions and broad eplet tallies contributing little. Overall, in this general cohort with a pooled endpoint, molecular mismatch metrics did not materially improve global discrimination, although model-agnostic xAI revealed structured, low-amplitude patterns that motivate targeted follow-up.
Department
Computer Science
Major
Computer Science
First Advisor
Dr. Robert Green
First Advisor Department
Computer Science
Second Advisor
Dr. Christopher Rump
Second Advisor Department
Applied Statistics and Operations Research
Publication Date
Fall 12-3-2025
Repository Citation
Shaik, Sourab, "Explainable AI for Liver Transplant Survival Prediction: Integrating Immunological Mismatch Features" (2025). Honors Projects. 1156.
https://scholarworks.bgsu.edu/honorsprojects/1156
Included in
Artificial Intelligence and Robotics Commons, Data Science Commons, Immunology and Infectious Disease Commons, Survival Analysis Commons, Vital and Health Statistics Commons