Nature Methods最新文献

The in silico labeling prediction of organelle fluorescence from label-free microscopy images has the potential to revolutionize our understanding of cells as integrated complex systems. However, out-of-distribution data caused by changes in the intracellular organization across cell types, cellular processes or perturbations can lead to altered label-free images and impaired in silico labeling. Here we demonstrate that incorporating biological meaningful cell contexts, via a context-dependent model that we call CELTIC, enhanced in silico labeling prediction and enabled the downstream analysis of out-of-distribution data such as cells undergoing mitosis and cells located at the edge of the colony. These results suggest a link between cell context and intracellular organization. Using CELTIC to generate single-cell images transitioning between different contexts enabled us to overcome intercell variability toward the integrated characterization of organelles' alterations in cellular organization. The explicit inclusion of context has the potential to harmonize multiple datasets, paving the way for generalized in silico labeling foundation models.

Imaging-based spatial transcriptomics methods allow for the measurement of spatial determinants of cellular phenotypes but are incompatible with random barcode-based clone-tracing methods, preventing the simultaneous detection of clonal and spatial drivers. Here we report SpaceBar, which enables simultaneous clone tracing and spatial gene expression profiling with standard imaging-based spatial transcriptomics pipelines. Our approach uses a library of 96 synthetic barcode sequences that combinatorially labels each cell. Thus, SpaceBar can distinguish between clonal dynamics and environmentally driven transcriptional regulation in complex tissue contexts.

Cancers differ in how they spread. These routes of metastatic dissemination can be reconstructed from tumor sequencing data, but current reconstruction methods scale poorly or rely on assumptions that do not reflect known biology. Metient overcomes these limitations using gradient-based, multiobjective optimization to generate multiple hypotheses of metastatic spread that are rescored using independent genetic distance and organotropism data. Unlike current methods, Metient can be used with both clinical sequencing data and barcode-based lineage tracing in preclinical models. Here, applied to data from 167 patients and 479 tumors, Metient identifies distinct trends of metastatic dissemination in melanoma, high-risk neuroblastoma and non-small cell lung cancer. Its reconstructions usually match expert analyses but Metient often finds other plausible migration histories, ultimately positing more polyclonal and metastasis-to-metastasis seeding than previously reported. Metient's reconstructions thus challenge existing assumptions about metastatic dissemination and offer insights into cancer type-specific patterns of metastatic spread.

Spatial omics face challenges in achieving high-parameter, multi-omics coprofiling. Serial-section profiling of complementary panels mitigates technical trade-offs but introduces the spatial diagonal integration problem. To address this, here we present SpatialEx and its extension SpatialEx+, computational frameworks leveraging histology as a universal anchor to integrate spatial molecular data across tissue sections. SpatialEx combines a pretrained hematoxylin and eosin foundation model with hypergraph and contrastive learning to predict single-cell omics from histology, encoding multi-neighborhood spatial dependencies and global tissue context. SpatialEx+ further introduces an omics cycle module that encourages cross-omics consistency via slice-invariant mappings, enabling seamless integration without comeasured training data. Extensive validations show superior hematoxylin and eosin-to-omics prediction, panel diagonal integration and omics diagonal integration across various biological scenarios. The frameworks scale to datasets exceeding 1 million cells, maintain robustness with nonoverlapping or heterogeneous sections and support unlimited omics layers in principle. Our work makes multimodal spatial profiling broadly accessible.

During developmental processes such as embryogenesis, how a group of cells self-organizes into specific structures is a central question in biology. However, it remains a major challenge to understand and predict the behavior of every cell within the living tissue over time during such intricate processes. Here we present MultiCell, a geometric deep learning method that can accurately capture the highly convoluted interactions among cells. We demonstrate that multicellular data can be represented with both granular and foam-like physical pictures through a unified graph data structure, considering both cellular interactions and cell junction networks. Using this method, we achieve interpretable four-dimensional morphological sequence alignment and predict single-cell behaviors before they occur at single-cell resolution during Drosophila embryogenesis. Furthermore, using neural activation map and model ablation studies, we demonstrate that cell geometry and cell junction networks are essential features for predicting cell behaviors during morphogenesis. This method sets the stage for data-driven quantitative studies of dynamic multicellular developmental processes at single-cell precision, offering a proof-of-concept pathway toward a unified morphodynamic atlas.