EPFL Releases Largest Open Spatial Proteomics Image Collection

The spora collection, built with EMBL-EBI, gathers thousands of cancer tissue images to help researchers compare and train AI models across studies.

Updated on Nov. 18, 2026 in Cancer

Isometric editorial illustration of stacked rectangular slide blanks filled with grids of colored dots on a cream field, suggesting organized shared tissue data.
EPFL released spora, a 13,546-image open collection from 5,477 cancer patients, designed to let researchers compare spatial proteomics studies and train AI models. AI Illustration. Upload story photo >

EPFL has released spora, described as the world's largest open-access spatial proteomics tissue image collection, containing 13,546 multiplex images from 5,477 patients. The release was made in collaboration with EMBL-EBI.

Why it matters

Spatial proteomics datasets have been hard to combine because laboratories use different equipment, examine different proteins, and store results differently. A unified, open collection could make cross-study comparison and AI model development more practical for cancer researchers.

The collection spans 13,546 multiplex images from 5,477 patients, 14 tissue origins and five imaging technologies, built over two years of work. The extraction does not provide a comparison against other datasets, and the collection currently focuses mainly on cancer research.

The players

EPFL (École Polytechnique Fédérale de Lausanne)

A leading Swiss public research university based in Lausanne that conducts research across engineering, science and computing.

EMBL-EBI (European Molecular Biology Laboratory – European Bioinformatics Institute)

A major European institute that provides open biological data resources and bioinformatics services to the global research community.

The details

Researchers standardized protein annotations, matched medical information and identified individual cells so data from different studies could be compared. The collection retains links to the original studies and their annotations and provides tools to access the data, develop AI models and evaluate model performance across studies.

Timeline

  1. Spora was built over a two-year period leading up to its release.

  2. Spora was released in 2026, with an associated NeurIPS 2026 publication.

Field Context

Spora extends the open-data approach of EMBL-EBI's biological resources into spatial proteomics, where data had been fragmented across laboratories. The release positions open cancer imaging collections as a foundation for training and benchmarking AI models.

Open access to standardized tissue imaging may accelerate research into cancer diagnosis and treatment, though the extraction does not describe any direct effect on patient care. Readers should expect any clinical application to depend on further validated studies that use these tools.

The takeaway

Open, standardized research data can make it easier for scientists to compare studies and build AI tools that may ultimately help the cancer field. Following future releases from spora and its partners will show how far the collection expands.

Further reading

Explore more Cancer coverage to see how open research data is advancing the field.

Source note: This article includes information reported by AIhub.org connecting the AI community and the world. - Association for the Understanding of Artificial Intelligence.