An Interpretable Machine Learning Framework for Rare Disease: A Case Study to Stratify Infection Risk in Pediatric Leukemia

• Type of Material: Article
• Authors: Irfan Al-Hussaini, Emory University; Brandon White, Emory University; Armon Varmeziar, Emory University; Nidhi Mehra, Emory University; Milagro Sanchez, Emory University; Judy Lee, Children’s Healthcare of Atlanta; Nicholas P. DeGroote, Children’s Healthcare of Atlanta; Tamara P. Miller, Children’s Healthcare of Atlanta; Cassie S. Mitchell, Emory University
• Language: English
• Date: 2024-03-20
• Publisher: MDPI
• Publication Version: Final Published Version
• Copyright Statement: © 2024 by the authors.
• License: Creative Commons Attribution 4.0 International
• Final Published Version (URL): https://doi.org/10.3390/jcm13061788
• Title of Journal or Parent Work: Journal of Clinical Medicine
• Volume: 13
• Issue: 6
• Start Page: 1788
• Grant/Funding Agency: Georgia Institute of Technology; Chan Zuckerberg Initiative; National Science Foundation; Children’s Healthcare of Atlanta; Aflac Cancer and Blood Disorders Center
• Grant/Funding Information: This research was funded by the Georgia Institute of Technology President’s Undergraduate Research Award to N.M. and B.W.; NIH grant R21CA232249 to C.M.; Aflac Pilot Grant from Aflac Cancer and Blood Disorders Center, Children’s Healthcare of Atlanta to T.P.M and C.S.M.; National Science Foundation CAREER award 1944247 to C.S.M.; and the Chan Zuckerberg Initiative grant 253558 to C.S.M.
• Keywords: pediatric leukemia; machine learning; artificial intelligence; natural language processing; infection
• Research Categories: Developmental biology; Oncology

Abstract

Background: Datasets on rare diseases, like pediatric acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL), have small sample sizes that hinder machine learning (ML). The objective was to develop an interpretable ML framework to elucidate actionable insights from small tabular rare disease datasets. Methods: The comprehensive framework employed optimized data imputation and sampling, supervised and unsupervised learning, and literature-based discovery (LBD). The framework was deployed to assess treatment-related infection in pediatric AML and ALL. Results: An interpretable decision tree classified the risk of infection as either “high risk” or “low risk” in pediatric ALL (n = 580) and AML (n = 132) with accuracy of ∼79%. Interpretable regression models predicted the discrete number of developed infections with a mean absolute error (MAE) of 2.26 for bacterial infections and an MAE of 1.29 for viral infections. Features that best explained the development of infection were the chemotherapy regimen, cancer cells in the central nervous system at initial diagnosis, chemotherapy course, leukemia type, Down syndrome, race, and National Cancer Institute risk classification. Finally, SemNet 2.0, an open-source LBD software that links relationships from 33+ million PubMed articles, identified additional features for the prediction of infection, like glucose, iron, neutropenia-reducing growth factors, and systemic lupus erythematosus (SLE). Conclusions: The developed ML framework enabled state-of-the-art, interpretable predictions using rare disease tabular datasets. ML model performance baselines were successfully produced to predict infection in pediatric AML and ALL.