Peripheral Blood DNA Methylation–Based Machine Learning Models for Prediction of Knee Osteoarthritis Progression: Biologic Specimens and Data From the Osteoarthritis Initiative and Johnston County Osteoarthritis Project

• Published in: Arthritis & rheumatology (Hoboken, N.J.) Vol. 75; no. 1; pp. 28 - 40
• Main Authors: Dunn, Christopher M., Sturdy, Cassandra, Velasco, Cassandra, Schlupp, Leoni, Prinz, Emmaline, Izda, Vladislav, Arbeeva, Liubov, Golightly, Yvonne M., Nelson, Amanda E., Jeffries, Matlock A.
• Format: Journal Article
• Language: English
• Published: Boston, USA Wiley Periodicals, Inc 01.01.2023; Wiley Subscription Services, Inc
• Subjects: Accuracy; Arthritis; Biological Products; Biomarkers; Blood; Blood cells; Buffy coat; CpG islands; Deoxyribonucleic acid; Disease Progression; DNA; DNA Methylation; Epigenetics; Genomes; Humans; Knee; Knee Joint; Learning algorithms; Machine learning; Mathematical models; Osteoarthritis; Osteoarthritis, Knee - diagnostic imaging; Osteoarthritis, Knee - genetics; Pain; Pain - etiology; Peripheral blood
• ISSN: 2326-5191; 2326-5205; 2326-5205
• DOI: 10.1002/art.42316

Objective The lack of accurate biomarkers to predict knee osteoarthritis (OA) progression is a key unmet need in OA clinical research. The objective of this study was to develop baseline peripheral blood epigenetic biomarker models to predict knee OA progression. Methods Genome‐wide buffy coat DNA methylation patterns from 554 individuals from the Osteoarthritis Biomarkers Consortium (OABC) were determined using Illumina Infinium MethylationEPIC 850K arrays. Data were divided into model development and validation sets, and machine learning models were trained to classify future OA progression by knee pain, radiographic imaging, knee pain plus radiographic imaging, and any progression (pain, radiographic, or both). Parsimonious models using the top 13 CpG sites most frequently selected during development were tested on independent samples from participants in the Johnston County Osteoarthritis (JoCo OA) Project (n = 128) and a previously published Osteoarthritis Initiative (OAI) data set (n = 55). Results Full models accurately classified future radiographic‐only progression (mean ± SEM accuracy 87 ± 0.8%, area under the curve [AUC] 0.94 ± 0.004), pain‐only progression (accuracy 89 ± 0.9%, AUC 0.97 ± 0.004), pain plus radiographic progression (accuracy 72 ± 0.7%, AUC 0.79 ± 0.006), and any progression (accuracy 78 ± 0.4%, AUC 0.86 ± 0.004). Pain‐only and radiographic‐only progressors were not distinguishable (mean ± SEM accuracy 58 ± 1%, AUC 0.62 ± 0.001). Parsimonious models showed similar performance and accurately classified future radiographic progressors in the OABC cohort and in both validation cohorts (mean ± SEM accuracy 80 ± 0.3%, AUC 0.88 ± 0.003 [using JoCo OA Project data], accuracy 80 ± 0.8%, AUC 0.89 ± 0.002 [using previous OAI data]). Conclusion Our data suggest that pain and structural progression share similar early systemic immune epigenotypes. Further studies should focus on evaluating the pathophysiologic consequences of differential DNA methylation and peripheral blood cell epigenotypes in individuals with knee OA.