Journal articles, conference and workshop papers, and preprints. Citation counts live on Google Scholar, which stays more current than anything I would maintain by hand.
Machine Learning in Computational Biology (MLCB)2026Conference paper
Reconstructs the spatial arrangement of cells in a tissue from gene expression alone, by predicting the distance between every pair of cells and recovering coordinates from that matrix.
Machine Learning in Computational Biology (MLCB)2026Conference paper
Each cell receives a short tuple of integers for its own identity and for its niche, giving a vocabulary that holds across sections and lets a query sample be matched to a reference by comparing code distributions.
bioRxiv2026Preprint
A foundation model pretrained on 112 million human cells that represents genes, cells and their spatial neighbourhoods from a single set of weights.
ICML 2026 Workshop on Graph Foundation ModelsWorkshop
ST-JEPA turns a cell and its spatial neighbours into a token sequence and learns by predicting the masked parts in latent space, giving representations at gene, cell and neighbourhood resolution without hand-designed augmentations.
ICLR 2026 Workshop on Machine Learning for Genomics ExplorationsWorkshop
Vector quantisation of cell embeddings from a graph neural network turns a tissue into a sequence of tokens, which SQUINT then uses to fill in gene expression at spatial locations it has not seen.
bioRxiv2026Preprint
A spatial atlas of around five million human skin cells uncovers immune niches inside sebaceous and sweat glands that standard histopathology does not reveal, which expand in atopic dermatitis and psoriasis and are still present twelve weeks after treatment.
arXiv2025Preprint
Aligns haematoxylin and eosin tissue images with spatial transcriptomic profiles at several scales at once, so the shared representation reflects individual cells and their arrangement rather than only the whole image tile.
bioRxiv2025Preprint
Separates the gene expression a cell owes to its microenvironment from the expression intrinsic to it, so that a microenvironment can be perturbed in silico rather than only described.
Nature Genetics 57(4), 897–9092025Journal
A graph deep-learning method that identifies cell niches in spatial omics data and quantifies the communication pathways that define them.