Abstract
<jats:p>Background: Candidate tumour-cell biomarkers identified from bulk transcriptomes are frequently expressed in stromal or vascular compartments as well, so a bulk correlation between such a gene and a biological programme can reflect tumour-cell biology or co-variation with those compartments. FOXC1, a PAM50 basal-defining transcription factor with canonical vascular expression, has immune associations in Luminal A breast cancer that have been read as tumour-cell-intrinsic. Methods: We developed a compartment-adjustment framework - purity- and stroma-adjusted partial correlations benchmarked across candidate marker genes - and applied it across three Luminal A cohorts (TCGA, (n=571); METABRIC, (n=700); SCAN-B, (n=1,540); two sequencing platforms) and single-cell data (GSE176078; 100,064 cells). Results: FOXC1's apparent adaptive-immune and tertiary-lymphoid-structure coupling is a vascular readout: it attenuates under adjustment for leukocyte-adhesion endothelial markers but persists under a structural-only endothelial composite, marking immune-recruiting vasculature. Single-cell analysis localises FOXC1 to the vessel wall-malignant cells contribute (1.8%) of FOXC1 transcripts versus (90%) from endothelial and perivascular cells. Basal cytokeratins (KRT5, KRT14, KRT17) and TP63, unlike FOXC1, retain a residual population-level basal-lineage signal; an apparent survival advantage is largely age-explained. Conclusions: Bulk FOXC1 in Luminal A originates principally from stromal--vascular compartments. Compartment adjustment offers a candidate approach for interpreting bulk biomarkers in heterogeneous tissue.</jats:p>