A Staged Whole-Blood Transcriptomic Framework Identifies a Compact Myeloid-Lymphoid Activity Score in Systemic Lupus...

Peripheral-blood transcriptomic profiling can capture molecular heterogeneity in systemic lupus erythematosus (SLE), but discovery-stage signatures often show limited transportability across cohorts and validation layers. This study aimed to establish a staged whole-blood transcriptomic framework and to derive a compact, biologically interpretable activity score.

Public whole-blood bulk transcriptome cohorts were organised into discovery, public validation, and single-cell reference layers. Local orthogonal validation included a peripheral blood mononuclear cell (PBMC) reverse transcription quantitative PCR (RT-qPCR)/flow-cytometric cohort and an expanded whole-blood RT-qPCR validation set. Discovery-stage BloodGen3 profiling included 233 samples, comprising 170 SLE and 63 healthy controls, and endotype discovery was restricted to SLE samples. Candidate genes were compressed into two 6-gene panels, with final selection adjudicated through staged public validation.

Two working whole-blood endotypes were identified, characterised by lymphoid versus myeloid/neutrophil-inflammatory polarisation. Although pre6-any showed a marginal discovery-stage advantage, the predefined integrated public-stage adjudication favoured pre6-balanced (MMP9, MYL9, HAL, CTLA4, CD40LG, VPREB3), which was locked as the final panel. In the PBMC cohort, the locked score discriminated SLE from healthy controls (AUC 0.838) and high from low/moderate disease activity (AUC 0.719), with associations with SLEDAI, complement C3/C4, and monocyte subpopulations. In the expanded whole-blood validation set, the score showed SLE-versus-HC discrimination (AUC 0.888, 95% CI 0.821-0.954), high versus low/moderate activity discrimination (AUC 0.918, 95% CI 0.831-0.980), and association with SLEDAI (ρ = 0.819, p = 1.25 × 10^-15).

This staged framework yielded a compact myeloid-lymphoid activity score supported across public and local validation layers. The score should be interpreted as a research-grade relative activity score and warrants prospective evaluation in SLE.