W196 - Peripheral Immune Signatures and Patient Stratification in Osteoarthritis: Application of Machine Learning Approaches to High-dimensional Immune Cytometric Data

ahmad Lajam – Translational Immunology Institute; gladys Ang – Translational Immunology Institute; Martin Wasser – Translational Immunology Institute; Ying Ying Leung – Rheumatology and Immunology – Singapore General Hospital; Salvatore Albani – Translational Immunology Institute

Abstract Text: Osteoarthritis (OA) is a prevalent joint condition that is especially common among older people. It is recognised that OA pathogenesis involves a complex interplay of joint damage, poor cartilage regeneration, and inflammation. In addition to the lack of approved disease-modifying drugs available for OA, clinical management is complicated by variable disease progression and limited treatment response. This suggests disease heterogeneity with subtypes of different disease trajectories and treatment responsiveness. We employed mass cytometry time-of-flight (CyTOF) to investigate peripheral blood mononuclear cells (PBMCs) obtained from patients with knee OA (n = 57) and matched healthy controls (HCs) (n = 33). Using immune clusters derived from the high-dimensional CyTOF data, a combination of statistical and machine learning algorithms was used to extract disease-associated peripheral immune signatures for patient stratification. We identified an inflammatory OA subtype, characterised by enrichment of monocytes and effector T/T-like cells subsets, associated with increased pain, stiffness, poorer function. Furthermore, we also identified another group of patients exhibiting milder pain and stiffness where their immune profile suggests immune-modulation as seen in the increase in Treg and exhausted Th immune subsets in their peripheral immunome. In conclusion, our application of machine learning approaches to high-dimensional CyTOF data identified OA subtypes of distinct clinical phenotype and disease-associated peripheral immune signatures. By uncovering the underlying heterogeneity of peripheral immunome changes in OA, our findings may have applications as a precision medicine tool in OA management where individualized therapeutic decisions can be made according to patient’s peripheral immunophenotype.