A pan-cancer spatial atlas from Asia Pacific reveals recurrent spatial theratopes

Here we generated a pan-cancer spatial atlas of 400 tumours across 13 tissue types and 26 subtypes from the Asia-Pacific, a region under-represented in existing pan-cancer datasets. In a multimodal cohort we integrated single-cell spatial transcriptomics (1,999 genes) with more than 70 proteins measured in the same cells as built-in validation, alongside clinical data including ancestry, age, sex and survival. ASTRA1.0 revealed that each cancer subtype has distinct expression profiles and cellular compositions. Despite this heterogeneity, we identified eight recurrent spatial cellular neighbourhoods shared across tissue types, with macrophages as key organisers. We further developed a framework to quantify the co-localisation of FDA-approved drug targets with surrounding cell types. This framework classified tumours into four Spatial Theratope subtypes, each defined by a distinct signature of co-localised targets. The VEGF-A–PD-L1 pair, targeted by bevacizumab and atezolizumab, recurred across tumour types and neighbourhoods. Spatial Theratope subtypes and neighbourhood composition were each associated with progression-free survival (PFS) in the discovery cohort, and their combination was also associated with PFS in an independent cohort of 25 patients with hepatocellular carcinoma treated with atezolizumab plus bevacizumab. The ASTRA atlas provides a community resource for nominating rational, spatially informed combination therapies across cancers.