Emma K. Lundberg

A conditional latent diffusion transformer predicts immunofluorescence images of 12,800 human proteins from nucleus, ER and microtubule stains, outperforming PUPS and CELL-Diff, including on an unseen drug-treated cell line.

The paper uses single-cell spatial proteomics to show that metabolic enzyme expression and localization vary between cells, revealing autonomous metabolic heterogeneity.

The paper generates a CRISPRi perturbation cell atlas in human induced pluripotent stem cells, identifies novel regulators of pluripotency and metabolism, and proposes a machine learning framework for predicting perturbation recipes.

The paper constructs a cell type-resolved proteomic atlas of the human body using Deep Visual Proteomics, revealing protein abundance patterns and uncovering cancer-testis antigens in oocytes.

The paper analyzes how cell shape correlates with proteomic features across multiple cell lines using image-based spatial proteomics data.

The paper creates a high-resolution spatial map of cilia-associated proteins in the human fallopian tube using integrated transcriptomics and proteomics.

GenBio-PathFM is a foundation model for histopathology that achieves state-of-the-art performance with less training data by using a diverse data curation pipeline and a dual-stage learning strategy.

This study evaluates foundation models for predicting cellular responses to perturbations, finding that some pretrained embeddings improve predictions while others do not.

The paper uses Molecular Pixelation to analyze surface proteomic signatures of CAR T cells under acute and chronic stimulation, identifying spatial patterns linked to exhaustion.

CellFluxV2 is an image-generative model that predicts cell morphology changes under perturbations, achieving higher image fidelity than baselines and enabling batch-effect correction and cell-state interpolation.

This study maps spatial dynamics in early human heart development using single-cell and spatial transcriptomics, identifying cell states and novel features like cardiac fibroblasts and chromaffin cells.

The paper constructs a single-cell multiomics atlas of the human pancreas across development, health, and T2D, identifying plastic cell states and lineage trajectories.

The paper uses spatial proteomics to analyze primary cilia proteomes across cell lines, revealing cell-type specificity and single-cilia heterogeneity, and identifies novel cilia proteins and potential genetic variants.

This paper introduces RIBCA, a tool for automated cell type annotation in multiplexed tissue images, capable of annotating over 3 million cells across 40 human tissues without manual intervention.

FeatureForest uses foundation model embeddings with random forests to enable fast, accurate segmentation of complex biological images with minimal labeling.

PIP{Sigma}X provides an automated, user-friendly pipeline for analyzing multiplexed tissue images, enabling robust spatial analysis and integration with visualization tools.

This paper introduces mxDVP, a method combining high-plex imaging and mass spectrometry to profile pancreatic islets, identifying 12 endocrine subtypes and revealing spatial heterogeneity and rare cell states.

This study creates a high-resolution morpho-proteomic atlas of mitosis by decomposing it into 40 subsections and quantifying 4,350 proteins, revealing dynamic changes and links to cancer mutations.

This paper describes the construction and usage of the Human Reference Atlas (HRA) v2.0, which integrates multiscale data across 36 organs and provides tools for analysis and visualization.

SubCell is a deep learning model for fluorescence microscopy that captures cellular morphology and protein organization, outperforming existing methods and enabling a multimodal approach to gene function.

This study maps the spatial organization of the developing human lung using high-parametric protein imaging, revealing cell type composition and temporal changes across five developmental timepoints.

This paper introduces a simplified VQGAN training scheme for generating multi-channel organelle staining from bright field images, achieving top performance in a benchmark challenge.

The paper introduces the CM4AI project, which aims to create AI-ready datasets of human cell architecture using multimodal data from cell lines.

Bento enables subcellular analysis of spatial transcriptomics data by defining domains, annotating localization patterns, and quantifying gene colocalization, demonstrating its use on multiple datasets.

This study shows that Xist ribonucleoprotein complexes drive sex-biased autoimmunity in males, leading to autoimmune pathology and reprogramming of immune cells to resemble female patterns.

The paper describes a machine learning competition to develop automated image segmentation tools for anatomical structures across human organs, resulting in models that will be used in the Human Reference Atlas.

The BioImage Model Zoo provides a community standard, infrastructure, and open repository for sharing, testing, and deploying pre-trained deep learning microscopy models across multiple bioimage analysis tools.

The paper creates a comprehensive topographic atlas of early human lung development, identifying cell states, trajectories, and interactions to understand cell heterogeneity origins.

The paper integrates protein imaging and interaction data to create a unified map of eukaryotic cell structure, revealing new subcellular systems and protein associations.

Pairing AI-driven image analysis with automated laser microdissection and sensitive mass spectrometry yields proteomes of single cells or nuclei chosen by their morphology, with tissue position retained.