Bulk Transcriptomic Inference in Vestibular Schwannoma is Sensitive to Estimated Cell Composition: Cross-Cohort Analysis and Single-Cell Lineage Annotation.

Authors

Mingyang Li, Qingyi Huo, Zhimin Wu, Lixin Huang, Tao Sun, Yibo Zhao, Qingtian Liang, Yunmou Ou, Peixuan Yang, Shunyan Zhu, Yeqing Xie, Chichen Liu, Hui Wang

Published in

Journal of molecular neuroscience : MN. Volume 76. Issue 4. Nov 10, 2026. Epub Nov 10, 2026.

Abstract

Bulk transcriptomic differences between vestibular schwannoma (VS) and nerve reflect both gene-expression differences and tissue composition. We evaluated composition sensitivity in GSE39645 (31 VS; nine nerve/Schwann-cell controls), compared a fixed 346-gene set with GSE108524 (27 VS, including 10 sporadic and 17 NF2-related cases; four nerve controls), and used GSE230375 for deconvolution and donor-level pseudobulk. Tumor-only GSE216783 supplied a second reference, while GSE250061 supported descriptive broad-lineage annotation. In the prespecified group-plus-batch model with five MuSiC composition principal components, 326 of 1,410 baseline differentially expressed genes retained both false-discovery rate < 0.05 and absolute log2 fold change > 1 thresholds; 1,084 lost at least one threshold, including 117 with reversed direction, and 405 were adjusted-only threshold calls. The primary design had group variance-inflation factor (VIF) 8.747; 303/500 sample-stratified bootstrap draws exceeded VIF 10 but remained full rank and were retained. Across the fixed 346 genes, direction concordance was 93.4% for all VS, 94.5% for sporadic cases and 93.6% for NF2-related cases. All-VS Pearson correlation was 0.883 (95% percentile interval 0.834-0.897); intervals were conditional on fixed references, annotations and gene set. A 20 + 20 + 20-repeat simulation showed fewer conditional beta-zero composition-only calls but greater expression-only standard errors and root-mean-square coefficient error, with VIF > 10 in 4/20 estimated-MuSiC and 3/20 oracle strong-scenario fits; all fits remained full rank. These analyses quantify model sensitivity; neither broad-lineage assignment nor simulation establishes mechanism, biomarkers or therapeutic targets.

PMID:
43007562
Bibliographic data and abstract were imported from PubMed on 11 Nov 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.