Interactive Visualization of Spatial Omics Data
Spatial omics technologies are revolutionizing the study of tissue architecture and cellular communication. The data acquired often consist of molecular expression data together with high-resolution images. The established data format for single-cell molecular expression profiles in Python is AnnData. High-resolution images, however, have very different data representations, mainly due to differences in experimental technology, data acquisition and pre-processing. Given this duality in terms of formats for data acquired with spatial omics technologies, it would be highly beneficial for analysts to be able to interactively visualize and explore analysis results of these two data representations in an integrative way.
To this end, the project will develop a napari plugin for the interactive visualization of spatial omics data. The napari plugin will take two inputs, including AnnData object—to store molecular profiles and annotation results, such as clustering and dimensionality reduction results, gene expression profiles and annotations, multimodal data etc.; and an image-data object—working across different formats including Squidpy’s ImageContainer, Starfish’s ImageStack and others. The napari plugin will provide methods to explicitly cast and align the AnnData observation coordinates to the image-data coordinates, whatever the representation is used at hand. Such methods will implement basic operations to handle coordinate transformation (e.g. shifting, scaling etc.) and reordering of dimensions and channels in a consistent way, in order to align the two data representations and enable the joint interactive visualization of AnnData and image data. The napari plugin will be easily extendable by developers to create a new loading method for the image format at hand.
This project plans to increase the visualization functionality of the current implementation such as pie-charts to visualize cell mixtures and graph-based plotting functionalities. Finally, a refactoring of napari widgets available in the current implementation will be performed, to use more extensively the native napari’s magicgui. The project aims to unify the integrative visualization of spatial omics data, effectively providing one of hopefully many upcoming bridges between the single-cell analysis community and the image analysis community in Python.