Research by Kleio-Maria Verrou et al. at the Nikolaou lab presents a novel machine-learning method to distinguish autoimmune from infectious diseases from whole-blood transcriptomes.
Despite the distinct priming events, most pathogenic responses against foreign antigens, as in infections, or self-antigens, as in autoimmune diseases, engage overall similar immunologic components. To date, there are no established molecular markers that can distinguish between anti-self and anti-foreign immune responses. As a result, diagnostic dilemmas often exist during the initial diagnosis of an autoimmune disease versus infection but also, more importantly, when a patient under immunosuppressive treatment for a systemic autoimmune disease presents with signs and symptoms that could reflect either an infection or disease flare, making clinical decisions challenging.
In this work, a collaboration between the Nikolaou and Kollias labs with the lab of Petros Sfikakis at the Medical School of NKUA, we have developed a novel preprocessing method for the analysis of bulk transcriptomic profiles. We then implemented classification modeling on the largest -to date- compiled dataset, comprising 594 whole-blood samples from various autoimmune diseases and 615 samples from viral, bacterial, and parasitic infections. Our best model distinguished autoimmune versus infectious diseases with 98% accuracy. Moreover, through feature selection, we were able to identify a small set of 24 ultra-discriminative genes, which we propose to use as a signature that could form the basis of a clinical diagnostic panel.
Our findings provide a new mechanistic understanding of the pathogenic autoimmune response and a proof-of-concept approach using a small set of transcripts in a single blood sample for the differential diagnosis of patients presenting with inflammatory disorders.
Kleio-Maria Verrou, Nikolaos I. Vlachogiannis, Argyrios N. Theofilopoulos, Maria Tektonidou, Georgios Kollias, Christoforos Nikolaou and Petros P. Sfikakis. (2025) Machine learning-based identification of a transcriptomic blood signature discriminating between systemic autoimmunity and infection.
Med, 100840, Sep5: https://doi.org/10.1016/j.medj.2025.100840