Modern single-cell technologies measure the activity of thousands of genes in each individual cell. This reveals how heterogeneous cells are, even within a single organ, and how they change in disease. Analyzing the resulting large-scale datasets has so far required programming skills, expertise in machine learning, and substantial computing resources.
Leveraging Cell Atlases: Mapping New Cells onto Established References
ArchMap works much like a digital map service: Researchers upload their data and select a suitable cell atlas, a reference map that describes which cell types are present in a given organ.
Using machine learning, ArchMap maps each cell onto this reference and predicts its cell type. The platform also estimates how uncertain each prediction is. A poor fit to the reference may indicate an unusual or disease-associated cell state – but it may also be caused by technical factors. The platform therefore lets users examine these cells in more detail directly.
"Cell atlases contain a wealth of knowledge. With ArchMap, we are making this knowledge accessible to researchers who are not specialists in programming or machine learning," says Prof. Dr. Fabian Theis, Director of the Computational Health Center at Helmholtz Munich and Chair of Mathematical Modelling of Biological Systems at the Technical University of Munich (TUM).
Research in Practice: Aberrant Cells Identified in Pulmonary Fibrosis
The team demonstrated ArchMap in practice using idiopathic pulmonary fibrosis, a severe disease characterized by progressive scarring of the lung. The researchers compared cells from patients with the Human Lung Cell Atlas, which contains cells from healthy donors.
Most cells were reliably assigned. However, a specific population of connective tissue cells - known as adventitial fibroblasts –, deviated markedly from their healthy counterparts. Further analysis revealed hallmark features of fibrosis in these cells, including elevated expression of CTHRC1, an established marker of fibrotic fibroblasts.
Collaborative Research, Secure Data
Within ArchMap, results can be shared with selected collaborators and analyzed jointly. All data are encrypted in transit and at rest. The platform can also be deployed on a user's own infrastructure.
ArchMap already hosts numerous validated atlases covering a range of organs, including official Human Cell Atlas references. Researchers can also contribute their own atlases.
Detecting Disease Before Symptoms Appear
Many diseases begin with changes in individual cells long before symptoms emerge. Understanding these changes can enable earlier diagnosis and more targeted treatment.
Until now, the knowledge contained in large cell atlases has been largely accessible only to specialized computational labs. ArchMap opens it up to laboratories and clinics that generate their own samples but lack dedicated bioinformatics teams.
As more researchers align their data to the same validated references, results become more comparable and disease-relevant cell states can be identified more rapidly.
"Reference-based analysis is revolutionizing the way we interpret single-cell data. Instead of working everything out manually, we simply map new data onto a suitable reference and immediately see which cell types are present in the dataset. What used to take months can now be done in just a few hours," says Dr. Malte Lücken, Group Leader at the Institute of Computational Biology and the Institute of Lung Health and Immunity at Helmholtz Munich, and member of the German Center for Lung Research (DZL).
This lays important groundwork for the long-term development of new diagnostic and therapeutic approaches, for example in chronic lung diseases.
Free Software: www.archmap.bio
Original Publication:
Lotfollahi et al., 2026: ArchMap: A web-based platform for reference-based analysis of single-cell datasets. DOI: Nature Genetics. DOI: 10.1038/s41588-026-02756-y
