Integrative agent-based modeling of cutaneous lupus: single-cell benchmarking and in silico evaluation of B-cell therapies

Agent-based models (ABMs) offer a powerful framework for representing emergent immune dynamics, yet their integration with single-cell datasets and therapeutics simulation remains limited. Here, we presented an integrative, single-cell informed systems immunology modeling framework built on universal immune system simulator (UISS) to mechanistically represent autoimmune skin inflammation specifically systemic and cutaneous lupus erythematosus. The framework incorporates (i) data-driven initialization from lesion single-cell RNA sequence cell fractions, (ii) modular immune processes coupling type I interferon signaling, ultraviolet induced keratinocyte injury, and (iii) a digital patient generation pipeline that captures inter-individual variability through stochastic parameter perturbation and quality-controlled cohort assembly. Using this platform, we constructed a cohort of 50 digital patients initially and implemented modular therapy components for B-cell Activating Factor (BAFF) neutralization (belimumab) and anti-CD20 B-cell depletion (rituximab). In silico, pre–post treatment experiments revealed heterogeneous serologic responses, modest median reductions in immunoglobulin G antibodies that target the Ro/SSA protein (Ro-reactive IgG), and strong correlations between plasmablast dynamics and circulation autoantibodies, highlighting the mechanistic constraints on B-cell targeted therapies. Overall, our results demonstrated how single-cell benchmarking, rule-based immune modeling, and digital patient ensembles can be integrated into a generalizable computational workflow for studying autoimmune pathophysiology and simulating therapeutic mechanisms to explore mechanistic determinants of disease progression and therapeutic response.