Multiomics combined with machine learning identifies cytoskeleton-related molecular signatures in peripheral immune cells of rheumatoid arthritis

Rheumatoid arthritis (RA) is an autoimmune disease characterized by profound cytoskeletal dysregulation in immune cells, yet the potential regulatory molecules remain poorly understood. We integrated proteomic and phosphoproteomic profiling of peripheral blood mononuclear cells from 96 RA patients and 90 healthy controls with an external DNA microarray cohort. Bioinformatics analysis and machine learning models were applied to identify cytoskeleton-associated molecular signatures, upstream transcription factors and kinase activities. Immune cell composition effects were assessed using computational deconvolution. In silico drug prediction and external dataset validation were further conducted to explore potential therapeutic candidates targeting key regulators. In addition, experimental validation including quantitative PCR and Western blotting was performed to assess cytoskeletal molecules. We identified FLNA S2152 as a key phosphorylation feature, while STAT1, PML, CBFB and RAD21 were identified as important upstream transcriptional regulators. Kinase network analysis further suggested the activation of PKC, CAMK, and AKT signaling pathways upstream of FLNA S2152 phosphorylation. Virtual drug prediction suggested that raltitrexed may be a potential therapeutic candidate by suppressing STAT1 in RA. External dataset analysis supported a partial association between cytoskeleton-related features and treatment response. qPCR and Western blot analyses further indicated the dysregulation of cytoskeleton protein expression in RA samples. This study reveals coordinated transcriptional and post-translational alterations associated with cytoskeletal remodeling in RA immune cells. These findings not only provide new insights into the cytoskeletal dysregulation but also suggest potential therapeutic hypotheses for RA.