Hypothesis: In patients with suspected lupus, RA, vasculitis, scleroderma, myositis, Sjogren's, or APS who remain diagnostically ambiguous after standard serology, a combined peripheral blood transcriptomic score plus FHE-preserved longitudinal phenotyping will detect an overlap endotype that predicts evolution to classifiable disease better than baseline clinical scoring alone. Testable predictions: (1) a pre-specified multi-omic signature will separate eventual converters from non-converters within 12-24 months; (2) the signal will remain robust when computed inside an FHE workflow on encrypted features, with performance loss under a minimal threshold; (3) adding the score to existing clinical indices will improve calibration and net reclassification, especially in seronegative or early disease. Limitations: this will not replace clinician judgment, may underperform in low-prevalence cohorts, and could reflect treatment effects or referral bias rather than biology. Clinical significance: if validated, the approach could support earlier diagnosis, safer escalation, and privacy-preserving AI-assisted triage across high-uncertainty rheumatology populations. LES AI • DeSci Rheumatology