Spatial proteomics is limited by detection sensitivity, multiplexing and multimodal integration, leaving a gap between discovery and clinical assays. Here we present protein and nucleic acid serial tyramide amplification (PASTA), using horseradish peroxidase-mediated oligonucleotide deposition and cyclical imaging for high-plex, multimodal spatial profiling. Compatible with conjugated antibodies and in situ hybridization, PASTA enables simultaneous protein and RNA codetection from formalin-fixed, paraffin-embedded samples, providing a cost-effective bridge from discovery to clinical validation.

Biological generative models can predict biological functions without task-specific training data but often under-perform specialized models. This is due to a fundamental 'alignment gap', where the rules learned during unsupervised training are not related to the function of interest. Here we demonstrate how to provide task-specific information without losing the general knowledge learned during pretraining by using direct preference optimization to align a structure-conditioned protein language model to preferentially generate stable protein sequences. Our aligned model, ProteinDPO, achieves stability prediction competitive to task-specific models and consistently outperforms unsupervised and fine-tuned versions of the model. Notably, ProteinDPO generalizes beyond its training data to enable stabilization and improved binding affinity prediction of large multichain protein complexes. When applied to stabilization of the hemagglutinin trimer, a primary component of influenza vaccines, ~80% of designs achieve increased or similar stability compared with the native hemagglutinin and up to 32 °C improvements from recently emerged mammalian strains. Our results demonstrate how to augment generative models with biophysical information and, more broadly, provide a general framework for the alignment of biological foundation models.

Artificial intelligence (AI) is rapidly transforming stem cell and developmental biology, offering new strategies to analyze, interpret and optimize complex, dynamic systems such as organoids and stem cell-derived embryo models. In this Perspective, we chart the integration of AI into image-based analysis of stem cell systems, highlighting how deep learning, convolutional neural networks and emerging foundation models enable automated classification, segmentation and phenotyping at increasing scale and precision. We showcase applications in phenotyping, drug screening and mechanistic discovery, including real-time fate prediction and the identification of hidden morphological signatures linked to differentiation and disease. Practical challenges, including limited annotated data, model interpretability and live imaging constraints, are examined alongside future opportunities, such as multimodal integration, real-time experimental steering and protocol optimization. Altogether, we argue that AI is not merely an analytical tool, but a discovery engine that enhances reproducibility, accelerates insight and brings us closer to a mechanistic understanding of self-organization in complex stem cell-derived systems.

The difficulty of particle picking in cryo-electron tomography remains a barrier to routine in situ structure determination. Machine learning is well suited to overcome this bottleneck with efficient algorithms that generalize across molecular species. To spur new algorithm development, we held a 3-month Kaggle challenge that tasked contestants with annotating five molecular species across hundreds of experimental tomograms. Here we analyze the results of this competition, which successfully engaged >1,000 participants and delivered particle pickers that outperformed existing state of the art. Systematic comparisons of the contestants' submissions revealed the tolerance of subtomogram averaging to moderate but not severe over-picking and underscored the need for more robust measures of annotation quality. The winning models also highlighted the importance of data augmentation to overcome limited training data. All competition tomograms along with the ground truth and winning teams' annotations have been released on the CryoET Data Portal as a resource to benchmark current and future particle picking algorithms.

Neurons interact in networks distributed throughout the brain. While much effort has focused on whole-brain calcium imaging, advances in genetically encoded voltage indicators raise the question of whether it might be possible to image neuronal voltage across entire brains. Achieving this requires a microscope with high volumetric imaging rates and signal-to-noise ratio. Here we present a remote-scanning light-sheet microscope capable of imaging genetically encoded voltage indicator-expressing neurons distributed throughout much of the brain of larval zebrafish at a volumetric rate of 200.8 Hz. We measured voltage traces from approximately one-quarter of all brain neurons. We found that neurons firing at different times during a sequence occupied different locations: visually evoked sequences mapped across the optic tectum, whereas stimulus-independent bursts were mapped across the cerebellum and medulla. Imaging voltage of neurons distributed in many brain regions may open new frontiers for understanding fundamental neural system operations.

Mechanical force-driven signaling has emerged as a key regulator of cell-cell interactions (CCIs), which can enhance immune cell function. However, current biochemical approaches for studying CCIs offer minimal direct control over cellular bulk phenotypes, while synthetic biomaterial systems fail to mimic the dynamic complexity of cells. Here we introduce kpiCells, a biomaterial-based platform that uses a biomimetic membrane-endoplasmic architecture to enable finely tuned phenocopying of cellular states via modular mechanical, chemical and topographical inputs. We demonstrate that kpiCells can engage in physiological CCIs and reproduce critical subcellular features. In T cell systems, kpiCells enable integrated interrogation of afferent mechanosensing pathways and efferent force-exertion pathways, and support measurement of piconewton-scale forces at individual T cell antigen receptors as well as single cell-cell force fingerprints that define activation thresholds. This work establishes kpiCells as a bionic model that enables synthetic material design with the level of functional complexity approaching living cell systems.