Page 206 - 2019_01-Haematologica-web

J.S. Gandelman et al.
of the computational and decision-tree analyses. We were unable to analyze whether clusters predicted response to therapy, as this was an observational cohort in which patients were on any systemic therapy at study entry. Thus, treatment response is an outcome of interest in assessing the utility of machine learning for chronic GvHD outcome stratification. An external validation cohort is pending for this analysis. External validation of machine-learning approaches is the gold standard, and external validation is necessary prior to clinical application of the findings.
These results have the potential to be applied to stratify risk in the clinical setting, enhance the current chronic GvHD classification system, refine inclusion criteria for phase 2 trials, and guide biomarker discovery for more specific therapeutic targets. The distillation of machine- learning knowledge into a decision tree increases the fea- sibility of clinical application of the clusters. However, the clusters have not been externally validated, and this step should be explored before clinical application.
Lastly, this a flexible machine learning-inspired work-
flow with numerous potential applications. The stability of the clusters suggests that this approach will be highly useful in revealing groups not only for this disease but for others that have complex phenotypes. Although the end- point for this analysis was overall survival, this workflow could be applied to explore whether clusters of patients differ in treatment response or composite chronic GvHD endpoints, such as failure-free survival. Additionally, this workflow has the potential to be applied to other human diseases with complex classification systems such as myelodysplastic syndrome and brain tumors. This approach may change the classification of human disease by revealing otherwise unapparent, clinically relevant patterns.
Acknowledgments
This work was supported by funding from The Vanderbilt Medical Scholars Program and Vanderbilt Ingram Cancer Center. We thank Dr. Yan Guo for helpful discussions, Allison Greenplate and Benjamin Reisman for helpful insight on figure design, and Dr. Paul Martin for his critical review of the manuscript.
References
1. Arora M, Cutler CS, Jagasia MH, et al. Late Acute and chronic graft-versus-host disease after allogeneic hematopoietic Cell Transplantation. Biol Blood Marrow Transplant. 2016;22(3):449-455.
2. Socié G, Ritz J. Current issues in chronic graft-versus-host disease. Blood. 2014;124 (3):374-384.
3. Filipovich AH, Weisdorf D, Pavletic S, et al. National Institutes of Health Consensus Development Project on Criteria for Clinical Trials in Chronic Graft-versus-Host Disease: I. Diagnosis and Staging Working Group report. Biol Blood Marrow Transplant. 2005;11(12):945-956.
4. Jagasia MH, Greinix HT, Arora M, et al. National Institutes of Health Consensus Development Project on Criteria for Clinical Trials in Chronic Graft-versus-Host Disease: I. The 2014 Diagnosis and Staging Working Group report. Biol Blood Marrow Transplant. 2015;21(3):389-401.e381.
5. Cooke KR, Luznik L, Sarantopoulos S, et al. The biology of chronic graft-versus-host dis- ease: a task force report from the National Institutes of Health Consensus Development Project on Criteria for Clinical Trials in Chronic Graft-versus-Host Disease. Biol Blood Marrow Transplant. 2017;23(2): 211-234.
6. Yu J, Storer BE, Kushekhar K, et al. Biomarker panel for chronic graft-versus- host disease. J Clin Oncol. 2016;34(22):2583- 2590.
7. Hartwell MJ, Ozbek U, Holler E, et al. An early-biomarker algorithm predicts lethal graft-versus-host disease and survival. JCI Insight. 2017;2(3):e89798.
8. Major-Monfried H, Renteria AS, Pawarode A, et al. MAGIC biomarkers predict long- term outcomes for steroid-resistant acute GVHD. Blood. 2018;131(25):2846-2855.
9. PaczesnyS,KrijanovskiOI,BraunTM,etal. A biomarker panel for acute graft-versus- host disease. Blood. 2009;113(2):273-278.
10. Inamoto Y, Martin PJ, Storer BE, et al. Association of severity of organ involve-
ment with mortality and recurrent malig- nancy in patients with chronic graft-versus- host disease. Haematologica. 2014;99(10): 1618-1623.
11. Esteva A, Kuprel B, Novoa RA, et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542(7639):115-118.
12. Ting DSW, Cheung CY, Lim G, et al. Development and validation of a deep learn- ing system for diabetic retinopathy and related eye diseases using retinal images from multiethnic populations with diabetes. JAMA. 2017;318(22):2211-2223.
13. Wang L, Liang R, Zhou T, et al. Identification and validation of asthma phenotypes in Chinese population using cluster analysis. Ann Allergy Asthma Immunol. 2017;119(4):324-332.
14. Ehteshami Bejnordi B, Veta M, Johannes van Diest P, et al. Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast can- cer. JAMA. 2017;318(22):2199-2210.
15. Lee SI, Celik S, Logsdon BA, et al. A machine learning approach to integrate big data for precision medicine in acute myeloid leukemia. Nat Commun. 2018;9(1):42.
16. Deo RC. Machine learning in medicine. Circulation. 2015;132(20):1920-1930.
17. Diggins KE, Ferrell Jr PB, Irish JM. Methods for discovery and characterization of cell subsets in high dimensional mass cytometry data. Methods. 2015;82:55-63.
18. van der Maaten LHG. Visualizing high- dimensional data using t-SNE. J Mach Learn Res. 2008;9:2579-2605.
19. Saeys Y, Van Gassen S, Lambrecht BN. Computational flow cytometry: helping to make sense of high-dimensional immunolo- gy data. Nat Rev Immunol. 2016;16(7):449- 462.
20. Singal AG, Mukherjee A, Elmunzer BJ, et al. Machine learning algorithms outperform conventional regression models in predict- ing development of hepatocellular carcino- ma. Am J Gastroenterol. 2013;108(11):1723- 1730.
21. Chronic GVHD Consortium. Rationale and design of the chronic GVHD cohort study:
improving outcomes assessment in chronic GVHD. Biol Blood Marrow Transplant. 2011;17(8):1114-1120.
22. Inamoto Y, Pidala J, Chai X, et al. Joint and fascia manifestations in chronic graft-versus- host disease and their assessment. Arthritis Rheumatol. 2014;66(4):1044-1052.
23. Amir E-aD, Davis KL, Tadmor MD, et al. viSNE enables visualization of high dimen- sional single-cell data and reveals phenotyp- ic heterogeneity of leukemia. Nature Biotechnology. 2013;31(6):545-552.
24. Van Gassen S, Callebaut B, Van Helden MJ, et al. FlowSOM: using self-organizing maps for visualization and interpretation of cytometry data. Cytometry A. 2015;87(7):636-645.
25. Diggins KE, Greenplate AR, Leelatian N, Wogsland CE, Irish JM. Characterizing cell subsets using marker enrichment modeling. Nat Methods. 2017;14(3):275-278.
26. Diggins KE, Gandelman JS, Roe CE, Irish JM. Generating quantitative cell identity labels with marker enrichment modeling (MEM). Curr Protoc Cytom. 2018;83: 10.21.11-10.21.28.
27. Palmer J, Williams K, Inamoto Y, et al. Pulmonary symptoms measured by the national institutes of health lung score pre- dict overall survival, nonrelapse mortality, and patient-reported outcomes in chronic graft-versus-host disease. Biol Blood Marrow Transplant. 2014;20(3):337-344.
28. Kitko CL, White ES, Baird K. Fibrotic and sclerotic manifestations of chronic graft ver- sus host disease. Biol Blood Marrow Transplant. 2012;18(1 Suppl):S46-S52.
29. Inamoto Y, Storer BE, Petersdorf EW, et al. Incidence, risk factors, and outcomes of scle- rosis in patients with chronic graft-versus- host disease. Blood. 2013;121(25):5098- 5103.
30. Inamoto Y, Martin PJ, Flowers MED, et al. Genetic risk factors for sclerotic graft-versus- host disease. Blood. 2016;128(11):1516-1524.
31. Arora M, Klein JP, Weisdorf DJ, et al. Chronic GVHD risk score: a Center for International Blood and Marrow Transplant Research analysis. Blood. 2011;117(24): 6714-6720.
196
haematologica | 2019; 104(1)