Lymphomas are clinically and genetically heterogeneous malignancies that arise from B or T‑cell precursors at different maturation stages. Although bulk omics identify driver mutations and support current classifications, patients with apparently similar genomic profiles often experience divergent outcomes. One likely cause of this variability is the spatial composition of the tumour micro‑environment (TME).

To characterise diverse TMEs, we built a 10x Visium spatial–transcriptomic atlas from 107 diagnostic lymph‑node biopsies that include Hodgkin lymphoma, diffuse large B‑cell lymphoma, follicular lymphoma, several rarer subtypes, and non‑tumour lymph nodes. Each sample was analysed with MEFISTO, a probabilistic framework that models spatial covariation to extract latent factors. Across the cohort, these factors consistently corresponded to micro‑regional gene programmes, such as germinal centre signalling, interferon response, hypoxia, and cellular respiration, together with additional pathways that were preferentially enriched in specific lymphoma subtypes.

Clinically relevant immune markers were also mapped. Aggressive tumours showed a tighter spatial association with CD86⁺ and CD80⁺ macrophages, whereas PD‑L1, CTLA‑4, and PD‑1 levels were comparable across entities, suggesting alternative immune‑modulatory circuits that bulk RNA‑seq cannot detect.

Our data show that biologically significant spatial programmes are present in lymphomas but remain undetectable by bulk RNA‑seq. This work therefore adds a crucial spatial dimension to lymphoma characterization.