spora: A Unified Multimodal Dataset and Platform for Spatial Proteomics
In the authors' words
Spatial proteomics (SP) captures the molecular composition and spatial organization of tissues at single-cell resolution, opening a direct view of the cellular ecosystems that drive disease. Protein panels, acquisition platforms, and tissue contexts differ from one cohort to the next, and a coherent picture of tissue biology can only emerge from models trained to bridge this heterogeneity. Foundation models offer the natural path forward, but their training depends on large, harmonized corpora that span platforms, panels, and cohorts. No public resource currently approaches that scale; available data sit in small, narrowly scoped releases with incompatible formats and conventions. We present , a large-scale multimodal SP dataset of 13,546 standardized multiplex images from 5,477 patients across 35 cohorts and five major SP technologies. Curated with biomedical experts, provides model training-ready formats together with cell and tissue segmentations, structured clinical metadata, and, where available, paired H&E enabling the training of virtual staining models. It is complemented by , a unified interface for tiling, sampling, and standardization across all cohorts and modalities, and , a benchmark suite spanning cell phenotyping, diagnostic and therapeutic prediction as well as virtual staining tasks across twelve cohorts that places current SP foundation models on common ground against task specific baselines.
In plain words AI summary
Spatial proteomics maps where proteins sit in tissue at single-cell detail. The authors built spora, a dataset of 13,546 standardized images from 5,477 patients across 35 cohorts. It includes cell maps, clinical records, and paired H&E slides for virtual staining training. Two tools, spora [io] and spora [bench], standardize data handling and test models on twelve cohorts.
Written by Kimi K2.5 using only the abstract above. Jev checked that every sentence is supported by the abstract sentences it cites, and removed any that were not. Point at a sentence to highlight its source. 1 sentence was removed for not being supported by the abstract.
Appeared: Tuesday, October 20. arXiv. Preprint, not yet peer-reviewed.
Authors' comment: Accepted to NeurIPS 2026