Fail-Closed Validation of Lineage-Associated Transcript-Count Diversity After IFN-β Stimulation in a Public Human PBMC Dataset

Shannon entropy of single-cell transcript counts reflects composition and sampling. We tested whether interferon-beta (IFN-β) stimulation was associated with donor-consistent diversity changes after employing technical controls. We reanalyzed GSE96583 peripheral blood mononuclear cells from eight donors with lupus, treating donors—not the 8949 target cells obtained from them—as inferential units. Analyses included sampling without replacement at 562 unique molecular identifiers, a 750-molecule sensitivity analysis, Miller–Madow correction, IFN-gene removal, and exact compositional decomposition. Mean paired-donor raw-count plug-in entropy increased by 0.138 bits in CD14+ monocytes and 0.329 bits in FCGR3A+ monocytes and decreased by 0.080 bits in natural killer cells. Directions persisted across prespecified robustness variants. Monocyte increases accompanied higher evenness and lower transcript dominance; natural killer cells showed the opposite pattern, without consistent richness changes. Adding IFN activity to donor-, condition-, and technical-covariate-adjusted models increased R2 by at most 0.0108. Entropy contributions strongly overlapped pseudobulk expression changes. In GSE194122, the implemented binary chromatin-accessibility entropy was determined by open-peak count, precluding biological cross-modality coupling claims. These findings support a lineage-divergent association within this single lupus cohort; independent replication is required before broader biological or diagnostic interpretation.