Evaluation of a Machine Learning-Based Prognostic Model for Unrelated Hematopoietic Cell Transplantation Donor Selection Ljubomir Buturovic, Jason Shelton, Stephen R. Spellman, Tao Wang, Lyssa Friedman, David Loftus, Lyndal Hesterberg, Todd Woodring, Katharina Fleischhauer, Katharine C. Hsu, Michael R. Verneris, Mike Haagenson, Stephanie J. Lee Biology of Blood and Marrow Transplantation Volume 24, Issue 6, Pages 1299-1306 (June 2018) DOI: 10.1016/j.bbmt.2018.01.038 Copyright © 2018 The American Society for Blood and Marrow Transplantation Terms and Conditions
Figure 1 The process used to define the set of variables and the model used for validation. Biology of Blood and Marrow Transplantation 2018 24, 1299-1306DOI: (10.1016/j.bbmt.2018.01.038) Copyright © 2018 The American Society for Blood and Marrow Transplantation Terms and Conditions
Figure 2 A graph of relevant statistics for a large collection of SVM classifiers developed for the HCT donor selection application. Each dot represents a classifier, which labels donors as Preferred (or, equivalently, “POS”, for Positive) and NotPreferred (or, equivalently, “NEG”, for Negative). The x-axis is the proportion of donors labeled Preferred (ie, POS) by the classifier. The y-axis is the survival benefit (difference in survival at 5 years) conferred by the donors, compared with survival of recipients who received HCT from NotPreferred donors. A clinically attractive classifier, selected for the validation, is labeled by a red arrow. It is defined as the classifier that maximizes clinical benefit while labeling at least 10% of donors as Preferred. Biology of Blood and Marrow Transplantation 2018 24, 1299-1306DOI: (10.1016/j.bbmt.2018.01.038) Copyright © 2018 The American Society for Blood and Marrow Transplantation Terms and Conditions
Figure 3 Survival of recipients of donors labeled Preferred and NotPreferred. The graph was produced using 10-fold cross-validation (HR, .43; 95% CI, .28 to .67; log-rank P < .001). Biology of Blood and Marrow Transplantation 2018 24, 1299-1306DOI: (10.1016/j.bbmt.2018.01.038) Copyright © 2018 The American Society for Blood and Marrow Transplantation Terms and Conditions
Figure 4 Survival of recipients of donors labeled Poor and NotPoor by the less stringent model, in 10-fold cross-validation (HR, .75; 95% CI, .61 to .91; log-rank P = .003.) Biology of Blood and Marrow Transplantation 2018 24, 1299-1306DOI: (10.1016/j.bbmt.2018.01.038) Copyright © 2018 The American Society for Blood and Marrow Transplantation Terms and Conditions
Figure 5 Validation Kaplan-Meier graph for the primary classification model (HR, 1.12; 95% CI, .72 to 1.72; log-rank P = .62). Biology of Blood and Marrow Transplantation 2018 24, 1299-1306DOI: (10.1016/j.bbmt.2018.01.038) Copyright © 2018 The American Society for Blood and Marrow Transplantation Terms and Conditions
Figure 6 Exploratory model validation results at 5 years (HR, 1.18; 95% CI, .94 to 1.48; log-rank P = .148). Biology of Blood and Marrow Transplantation 2018 24, 1299-1306DOI: (10.1016/j.bbmt.2018.01.038) Copyright © 2018 The American Society for Blood and Marrow Transplantation Terms and Conditions
Figure 7 . Primary classification model validation results for patients with AML (HR, 2.01; 95% CI, 1.22 to 3.3; log-rank P = .005) (A) and patients with ALL (HR, .42; 95% CI, .17 to 1.02; log-rank P = .049) (B). Biology of Blood and Marrow Transplantation 2018 24, 1299-1306DOI: (10.1016/j.bbmt.2018.01.038) Copyright © 2018 The American Society for Blood and Marrow Transplantation Terms and Conditions
Figure 8 . Primary classification model training (cross-validation) results for patients with AML (HR, .56; 95% CI, .28 to 1.09; log-rank P = .083) (A) and patients with ALL (HR, .37; 95% CI, .21 to .66; log-rank P < .001) (B). Biology of Blood and Marrow Transplantation 2018 24, 1299-1306DOI: (10.1016/j.bbmt.2018.01.038) Copyright © 2018 The American Society for Blood and Marrow Transplantation Terms and Conditions