Machine learning-driven stratification of Immune disorders using serum cytokine autoantibody profiles detected by cell-based multiplex flow cytometry 2259074

Introduction Impaired immune tolerance triggers aberrant or sometimes overshooting immune responses, contributing to disease manifestations and/or exacerbations. Notably, the prevalence of serum cytokine-specific autoantibodies (c-AAbs) has been documented in ‘healthy’ individuals as well as in association with acute and chronic states of immune disorders. Precise detection of c-AAbs, which can either neutralize, stabilize, or modulate the function of cytokines, requires sensitive assays that present the antigen in its native conformation. Methods A novel multiplex flow cytometry-based cellular assay (mFCCA) was devised based on a collection of stable HEK-293T single-cell clones expressing functional interferons and interleukins, for the assessment of c-AAbs in patients’ serum (n = 400), comprising controls, HIV-infected, and those with suspected hyperinflammatory, autoimmune, or malignant disorders characterized by laboratory parameters. The IgG immunoreactivity against 15 cytokine-expressing cell lines was evaluated in a multiplex format, alongside the parental cell line, with specific binding quantified as delta channel geometric mean fluorescence intensity. Subsequently, a cross-validated Random Forest algorithm was applied to discriminate between the five disease phenotypes and the controls. Model performance was gauged by the receiver operating characteristic curves. Results Distinct disease-specific c-AAb signatures were identified across patient groups. For instance, HIV patients exhibited significantly higher frequencies of anti-IFN-α2a, IL-1α, IL-2, and IL-5 autoantibodies compared with controls. The classification model demonstrated high accuracy, with area under the curve values ranging from 0.91 to 0.99, and correct assignment of 331 of 400 samples (83%) to their respective groups. Conclusion The high-throughput approach for screening c-AAb reactivities against surface-expressed cytokines via multiplexed flow cytometry enables rapid immunopathology classification and may assist targeted therapy. Funding Source The work was supported by the Federal State of Lower Austria under the Danube Allergy Research Cluster 2.0 (Danube-ARC 2.0) [grant no. FA624A0404 —Project#13] and the Medical University of Vienna.