FDZS: A Multi-Cohort Colorectal Liver Metastases Dataset for Spatial Transcriptomics and Proteomics Analysis

Adjuvant therapy (AT) improves outcomes for colorectal liver metastases (CRLM), yet predicting benefit, especially from adding targeted agents (Cetuximab or Bevacizumab) to chemotherapy, remains challenging. The spatial organization of the tumor microenvironment (TME) influences treatment response, yet its role in CRLM AT efficacy is unclear. We performed imaging mass cytometry (IMC) on 311 regions across tumor core, invasive margin, and peritumor (PT) tissues from 35 AT-treated CRLM patients. Systematic spatial proteomic analysis revealed distinct cholangiocyte-anchored niches within the PT region associated with recurrence-free survival (RFS): an "Immune-Rich" (PIR-Niche) enriched with B cells and CD8+ T cells, and a "Stromal-Metabolic" (PSM-Niche) dominated by CD163+ macrophages and Collagen-I+ stromal cells. The Immune-Stromal Ratio (ISR), a metric quantifying the relative abundance of these two niches, predicted RFS specifically in patients receiving combination therapy (p=0.044). To enhance clinical applicability, we developed SpMap, a deep learning tool inferring the ISR from routine H&E slides. The SpMap-derived ISR significantly predicted RFS in discovery (p=0.036) and independent test (n=95, p=0.023) cohorts, identifying high-ISR patients deriving significant benefit from combination therapy. Integrated spatial and single-cell transcriptomics confirmed distinct cellular compositions and revealed enrichment of angiogenesis pathways in PIR-Niches and mTORC1 signaling in PSM-Niches. Our work explores the spatial architecture within peritumor, captured by the ISR, as a potential predictor of adjuvant combination therapy response in CRLM. Furthermore, SpMap provides a H&E-based tool for patient stratification, while uncovering potential targetable pathways driving niche function.