spora: A Comprehensive Dataset for Spatial Proteomics Insights

Paper: spora: A Unified Multimodal Dataset and Platform for Spatial Proteomics

Authors: Benedikt von Querfurth, Johann Wenckstern, Eeshaan Jain, Lukas Klein, Phil Cheng, Yexiang Cheng, Cedric Vincent-Cuaz, Martina Haberecker, Pascal Frossard, Andreas Wicki, Olivier Michielin, Charlotte Bunne

Executive Summary

This research addresses a significant challenge in the field of spatial proteomics (SP), which studies the molecular makeup and arrangement of tissues at an individual cell level. Understanding this spatial organization is crucial for uncovering the cellular dynamics that contribute to various diseases. However, current research in this area faces limitations due to variations in protein panels, data acquisition platforms, and tissue contexts among different patient cohorts. These inconsistencies make it difficult to build comprehensive models that accurately reflect tissue biology.

To tackle these challenges, the authors introduce a new resource called spora, which is a large-scale, multimodal dataset encompassing 13,546 standardized multiplex images from 5,477 patients across 35 different cohorts and five major SP technologies. This dataset is significant because it is curated with input from biomedical experts, ensuring that it is robust and reliable for training models. The spora resource provides data in formats that are ready for model development, including detailed cell and tissue segmentations along with structured clinical metadata. Additionally, where applicable, it includes paired hematoxylin and eosin (H&E) stained images, which can assist in the training of virtual staining models.

The authors also present two complementary tools: spora [io], a unified interface designed for efficient data tiling, sampling, and standardization across various cohorts and modalities; and spora [bench], a benchmark suite that evaluates current SP foundation models against specific tasks. This suite covers cell phenotyping, diagnostic and therapeutic predictions, and virtual staining tasks across twelve cohorts, allowing researchers to compare their models on a common basis.

The implications of this research are broad. By providing a comprehensive, harmonized dataset, spora could facilitate the development of more accurate predictive models in spatial proteomics. This advancement may lead to improved diagnostics and therapeutic strategies in healthcare. However, it’s important to note that the results and capabilities discussed arise from simulations and benchmark evaluations rather than direct real-world applications.

Overall, the introduction of spora represents a critical step towards overcoming existing barriers in spatial proteomics research, potentially enabling significant advancements in understanding disease mechanisms at the cellular level.

Academic Abstract

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 $\texttt{spora}$, 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, $\texttt{spora}$ 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 $\texttt{spora [io]}$, a unified interface for tiling, sampling, and standardization across all cohorts and modalities, and $\texttt{spora [bench]}$, 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.