scVIVA: a probabilistic framework for representation of cells and their environments in spatial transcriptomics
Levy, N.; Ingelfinger, F.; Bakulin, A.; Cinnirella, G.; Boyeau, P.; Nadler, B.; Ergen, C.; Yosef, N.
Show abstract
Spatial transcriptomics provides a significant advance over studies of dissociated cells in that it reveals the environment in which cells reside, thus opening the way for a more complete description of their state and function. However, most current methods for embedding and discovery of cell states rely only on the cells own gene expression profile, thus raising the need for ways to account for the neighboring cells as well. Here, we introduce scVIVA, a deep generative model that leverages both cell-intrinsic and neighboring gene expression profiles to output stochastic embeddings of cell states as well as normalized gene expression profiles. We demonstrate that scVIVA produces informative fine-grained partitions of cells that reflect both their internal state and the surrounding tissue and that its generative model facilitates the testing of hypotheses of differential expression between tissue niches. We leverage these properties of scVIVA to uncover a spatially-restricted tumor-promoting endothelial population in breast cancer and niche-associated T cell states that are shared across multiple cancers. scVIVA is available as open source software within scvi-tools.org.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
Similar papers in this journal
- DeST-OT: Alignment of Spatiotemporal Transcriptomics Data 96%
- Learning multi-cellular representations of single-cell transcriptomics data enables characterization of patient-level disease states 95%
- Interpretable deep learning of label-free live cell images uncovers functional hallmarks of highly-metastatic melanoma 95%
Similar papers in this journal
- Neighborhood nonnegative matrix factorization identifies patterns and spatially-variable genes in large-scale spatial transcriptomics data 98%
- scAlign: a tool for alignment, integration and rare cell identification from scRNA-seq data 96%
- Smoother: A Unified and Modular Framework for Incorporating Structural Dependency in Spatial Omics Data 96%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.