Arthritis Care & Research · 2021 · 16 citations · 34 references
Life SciencesImmunologyInternal MedicineClinical ImmunologyInflammationRheumatologyAutoimmune BiomarkersAutoimmune Disease DiagnosticsAutoimmune DiseasesAutoimmune Rheumatic DiseasesInflammatory BiomarkersRheumatoid ArthritisBiomarkersDisease Phenotype IdentificationImmune Biomarker DiscoveryImmune Biomarker ProfilingImmune Signature ProfilingMachine Learning Biomarker DiscoveryRheumatoid Arthritis DiagnosticsSerological Profiling
OBJECTIVE: Native American (NA) populations have higher rates of rheumatic disease and present with overlapping disease symptoms and nontraditional serologic features, thus presenting an urgent need for better biomarkers in NA diagnostics. This study used a machine learning approach to identify immune signatures that more effectively stratify NA patients with rheumatic disease. METHODS: Adult NA patients with autoantibody-positive (AAB+) rheumatoid arthritis (RA; n = 28), autoantibody negative (AAB-) RA (n = 18), systemic autoimmune rheumatic disease (n = 28), arthralgia/osteoarthritis (n = 28), or polyarthritis/undifferentiated connective tissue disease (n = 28), and control patients (n = 28) provided serum samples for cytokine, chemokine, and AAB assessment. Random forest clustering and soluble mediator groups were used to identify patients and control patients with similar biologic signatures. The American College of Rheumatology criteria specific for systemic disease and RA identified differences in disease manifestations across clusters. RESULTS: Serum soluble mediators were not homogenous between different NA rheumatic disease diagnostic groups, reflecting the heterogeneity of autoimmune diseases. Clustering by serum biomarkers created 5 analogous immune phenotypes. Soluble mediators and pathways associated with chronic inflammation and involvement of the innate, B cell, T follicular helper cell, and interferon-associated pathways, along with regulatory signatures, distinguished the 5 immune signatures among patients. Select clinical features were associated with individual immune profiles. Patients with low inflammatory and higher regulatory signatures were more likely to have few clinical manifestations, whereas those with T cell pathway involvement had more arthritis. CONCLUSION: Serum protein signatures distinguished NA patients with rheumatic disease into distinct immune subsets. Following these immune profiles over time may assist with earlier diagnoses and help guide more personalized treatment approaches.
34
WGCNA: an R package for weighted correlation network analysis
Peter Langfelder, Steve Horvath · BMC Bioinformatics · 2008 · 29.1K citations · Full text
Pathologically expanded peripheral T helper cell subset drives B cells in rheumatoid arthritis
Deepak A. Rao, Michael F. Gurish, Jennifer L. Marshall et al. · Nature · 2017 · 1.1K citations · Full text