Disease maps are moving from tissue averages to cell-by-cell atlases

Maps instead of averages

The shared move in these preprints is away from grinding up tissue and averaging it. A new statistical framework called FICTURE2 models individual RNA molecules to infer transcriptional programs, avoiding the biases and computational overhead of cell segmentation or binning. Because it works at the molecule level, the authors report, it captures spatial dynamics from micron-scale subcellular localisation up to millimetre-scale transitions across a tissue, and scales to large spatial datasets. m:…0054

That matters because the biology being mapped is organised in space, not just in cell type. The ALS motor cortex preprint resolves about 600 postmortem samples into seven distinct transcriptomic subtypes through consensus clustering, describing a structured landscape of inflammatory, metabolic and neuronal programs. The subtypes track genetic factors such as C9orf72 status more strongly than disease duration, which is why the authors present the framework as a tool for patient stratification and target validation in ALS drug discovery. m:…0779

The practical payoff the authors claim for finer maps is better targeting: the ALS framework is presented as a tool to improve patient stratification and target validation in drug discovery, and the bone atlas as a map for drug discovery at the cartilage-bone interface. In both cases the map is useful only insofar as it changes which patients are grouped together and which targets are pursued, which is why the subtype and signalling details, rather than the headline diagnosis, carry the news. m:…0779 m:…7492

What the maps found

In bone, a single-nucleus multiome atlas of human subchondral bone overcomes the long-standing difficulty of studying mineralised osteocytes, and finds osteoarthritis progression organised around the cartilage-bone interface. The osteogenic compartment expands as disease worsens, and chondrocyte-to-osteocyte signalling via FN1 increases, activating TGF-beta/SMAD2/3 pathways in osteocytes and promoting matrix remodelling. The authors present the atlas as a high-resolution map for drug discovery aimed at that interface. m:…7492

In the oesophagus, a patient-derived organoid model stimulated with IL-13 reproduces the epithelial dysregulation of eosinophilic oesophagitis. Comparing single-cell RNA sequencing profiles from patient biopsies and organoids, the authors report that the model captures cell state heterogeneity, differentiation trajectories and inflammatory pathway activation, making it a scalable platform for testing epithelial-targeted therapies and mechanistic drivers of the disease. m:…7192

Read together, the three disease maps share a conclusion: the informative unit is smaller and more local than a diagnosis. ALS splits into seven molecular subtypes, osteoarthritis into a signalling axis at one tissue interface, and an oesophageal disease into a reproducible set of cell states in a dish. m:…0779 m:…7492 m:…7192

The limits of a map

All four are preprints, and none of them is a treatment. A subtype map does not by itself change ALS trial design until a therapy is tested within subtypes; an organoid that reproduces cell states still needs to predict drug response in patients; and a bone atlas points at an interface without yet showing that intervening there slows joint damage. m:…0779 m:…7492 m:…7192

The method preprint is the most portable of the four, because it is a tool rather than a finding: if molecule-resolved inference works across tissues and datasets, it lowers the cost of every future atlas. The disease preprints are the test of whether cheaper, finer maps actually change which targets get pursued, and that test will be decided by follow-up studies, not by the maps themselves. m:…0054