This six-part series demonstrates an end-to-end spatial transcriptomics workflow using the PortrAIgent platform to analyze TNBC and ER+ breast cancer tissues by integrating Visium data with a single-cell RNA reference. By combining cell type deconvolution, differential pathway profiling, and spatial distance mapping, the analysis successfully captures subtype-specific molecular hallmarks and reveals a compartmentalized tumor microenvironment bounded by a cancer core, CAF-rich stroma, and peripheral immune infiltration.
1. PortrAIgent Visium Breast Cancer Analysis: Data Overview & Preprocessing Setup
Dataset Inspection: Evaluates Visium data containing 15,611 spots and 18,714 genes with raw counts, alongside related single cell refernce containing 80,753 cells and 28,468 genes with log-normalized values in
Xand raw counts inraw.Sample & Metadata Overview: The Visium dataset comprises 6 patient samples (4 TNBC, 2 ER+). The scRNA reference includes 21 patients (10 TNBC, 11 ER+) with a 3-tier cell annotation (9 major cell types, 29 minor clusters, 49 subsets), clinical grade, treatment history, and ER/PR/HER2 IHC status.
Analytical Feasibility: Identifies that Visium lacks cell annotations and requires scRNA-guided deconvolution. Establishes TNBC vs. ER+ as the primary group comparison, notes that spatial analyses must be conducted per sample as all 6 sections lie on a single grid, and highlights the requirement to use raw counts from scRNA.
2. PortrAIgent Visium Breast Cancer Analysis: Batch Correction, Clustering & Spatial Annotation
Workflow Execution: Performs QC, normalization, HVG selection, PCA, Harmony batch correction, UMAP, Leiden clustering, Wilcoxon marker identification, and Squidpy spatial Leiden clustering.
Cell Annotation: Assigns cell types using marker genes and scRNA signature NNLS scores, projecting annotations onto UMAP and spatial tissue slides.
Key Results: Basal-like Cancer Epithelial cells (marked by KRT14, EGFR, KRT17) constitute 81% of TNBC sample 1142243F. CAFs (SPARC, COL1A2) account for 64% of TNBC sample CID4465. Plasmablasts (IGHG4, IGKC) and T-cell clusters (CD74, HLA-DRA, TRBC2) make up 36% of TNBC sample 1160920F. ER+ samples separate into distinct luminal clusters.
Biological Significance: Successfully reproduces basal-like features and high immune infiltration in TNBC, alongside luminal characteristics in ER+ breast cancer.
3. PortrAIgent Visium Breast Cancer Analysis: Cell Type Deconvolution via TACCO
Workflow Execution: Applies TACCO Optimal Transport deconvolution using 9 major cell types from the scRNA reference to estimate cell fractions per Visium spot. Generates cell type spatial maps, dominant cell maps, and group composition comparisons.
Key Composition Results: Cancer Epithelial cells are predominant in ER+ (0.60 fraction) compared to TNBC (0.15 fraction). TNBC shows higher proportions of T-cells (0.22 vs 0.07), Plasmablasts (0.08 vs 0.01), B-cells (0.05 vs 0.01), and Myeloid cells (0.11 vs 0.05).
Sample Highlights: Sample CID4290 is 97% tumor-dominant, CID4465 is CAF-dominant, and 1160920F exhibits localized plasmablast aggregation matching P1 clustering results.
Biological Significance: Confirms an immune-infiltrated microenvironment in TNBC versus a tumor-dense microenvironment in ER+ breast cancer.
4. PortrAIgent Visium Breast Cancer Analysis: Differential Expression & Pathway Enrichment
Workflow Execution: Evaluates gene expression differences between TNBC and ER+ groups using spot-level Wilcoxon tests and pseudobulk pydeseq2 analysis, followed by gseapy Enrichr ORA against MSigDB Hallmark gene sets.
Key DEG Findings: TNBC displays upregulation of basal markers including KRT14, KRT6B, KRT17, EGFR, SOX10, GABRP, FOXC1, and MKI67. ER+ displays upregulation of luminal markers including ESR1, GATA3, FOXA1, XBP1, AR, TFF1, and AGR2.
Enrichment Results: TNBC exhibits significant enrichment in Myc/E2F targets, G2-M checkpoint, interferon response, and glycolysis. ER+ is enriched in early/late estrogen response, androgen response, and fatty acid metabolism.
Biological Significance: Upregulated gene sets and pathways show complete alignment with PAM50 subtype classification and established breast cancer biology.
5. PortrAIgent Visium Breast Cancer Analysis: Spatial Pathway Activity Mapping via PROGENy
Workflow Execution: Calculates spot-level activity scores for Hallmark and PROGENy pathways using decoupler ULM. Produces patient-level heatmaps, group comparisons (TNBC vs. ER+), and spatial pathway maps.
Key Pathway Results: PROGENy pathway analysis demonstrates higher EGFR, MAPK, PI3K, JAK-STAT, NFkB, and TNFa activity alongside lower p53 activity in TNBC. Highly immune-infiltrated sample 1160920F exhibits peak JAK-STAT activity.
Hallmark Pathway Results: Hallmark analysis confirms elevated estrogen and androgen responses in ER+, with higher E2F/G2M targets and interferon response in TNBC.
Biological Significance: Recaptures EGFR/MAPK driver activity, TP53 mutation effects in TNBC, and JAK-STAT/NFkB inflammatory signaling in immune-infiltrated tumors.
6. PortrAIgent Visium Breast Cancer Analysis: Spatial Co-occurrence & TME Distance Gradient
Spatial Co-occurrence: Utilizes Squidpy co-occurrence analysis based on decoupler dominant cell types. Finds T-cell spots neighbor B-cell spots approximately 6x more frequently than expected by chance, Endothelial spots neighbor Plasmablasts and PVL, while Cancer Epithelium remains spatially segregated from immune cells.
Tumor Distance Gradient: Measures cell composition across distance intervals from the cancer core (spots with >= 70% NNLS Cancer Epithelial fraction). Cancer Epithelial presence drops from 82% at the core to 21% at 600–1,000 µm distance. Concurrently, CAFs increase from 11% to 32%, Myeloid cells from 5% to 18%, Plasmablasts from 0.3% to 9%, and Normal Epithelium rises to 15% beyond 1,000 µm.
Biological Significance: Demonstrates a compartmentalized tumor microenvironment featuring a tumor core, CAF-rich desmoplastic stroma, outer immune infiltrate, and surrounding normal tissue.
