Wen, Z.-H. | bioRxiv Science

bioRxiv2026

Branch-resolved IFN/STING-linked lesion-state architecture in psoriasis: multi-cohort derivation, held-out bulk replication and single-cell context analysis

Objective and designWe tested whether a frozen branch-resolved IFN/STING-linked lesion-state architecture derived from paired psoriasis transcriptomes would transport to an untouched bulk RNA-sequencing cohort and remain detectable within broad cell compartments. Material or subjectsSeven paired bulk cohorts formed the derivation backbone. GSE121212 provided held-out replication using 27 matched lesional/nonlesional pairs; treatment-facing cohorts and GSE228421 provided pharmacodynamic and cellular-context analyses. TreatmentNo intervention was administered by the authors; public datasets included ustekinumab-, etanercept-, secukinumab- and risankizumab-exposed samples. MethodsFrozen modules and branches were scored after cohort-wide gene standardization. GSE121212 counts underwent trimmed mean of M-values normalization and log2 counts-per-million transformation. Paired effects, upper-quartile STING-high contrasts, lesional Spearman coupling and prespecified rank concordance were evaluated. ResultsGSE121212 lesional-minus-nonlesional effects were 1.288 for STING-core, 1.478 for IFN-responsive activity and 1.188 for proximal-only STING; all lower 95% confidence limits exceeded zero. Coupling-vector concordance was strong (rho = 0.893, P = 0.0068), whereas STING-high effect concordance was moderate (rho = 0.536), yielding partial replication. ConclusionsHeld-out replication was partial: the IFN/STING lesion anchor and proximal/IFN-dominant coupling transported, while broader ordering did not fully transport. Transcriptomic findings do not establish biochemical STING activation, causality or clinical thresholds.

immunology↗

bioRxiv2021

Bfimpute: A Bayesian factorization method to recover single-cell RNA sequencing data

Single-cell RNA-seq (scRNA-seq) offers opportunities to study gene expression of tens of thousands of single cells simultaneously, to investigate cell-to-cell variation, and to reconstruct cell-type-specific gene regulatory networks. Recovering dropout events in a sparse gene expression matrix for scRNA-seq data is a long-standing matrix completion problem. We introduce Bfimpute, a Bayesian factorization imputation algorithm that reconstructs two latent gene and cell matrices to impute final gene expression matrix within each cell group, with or without the aid of cell type labels or bulk data. Bfimpute achieves better accuracy than other six publicly notable scRNA-seq imputation methods on simulated and real scRNA-seq data, as measured by several different evaluation metrics. Bfimpute can also flexibly integrate any gene or cell related information that users provide to increase the performance. Availability: Bfimpute is implemented in R and is freely available at https://github.com/maiziezhoulab/Bfimpute.

bioinformatics↗