STitch3D
We developed STitch3D, a deep learning-based method for 3D reconstruction of tissues or whole organisms. Briefly, STitch3D characterizes complex tissue architectures by borrowing information across multiple 2D tissue slices and integrates them with a paired single-cell RNA-sequencing atlas.
STitch3D enables two critical 3D analyses: First, STitch3D detects 3D spatial tissue regions which are related to biological functions, for example cortex layer structures in brain; Second, STitch3D infers 3D spatial distributions of fine-grained cell types in tissues, substantially improving the spatial resolution of seq-based ST approaches.
The output of STitch3D can be further used for various downstream tasks like inference of spatial trajectories, denoising of spatial gene expression patterns, identification of genes enriched in specific biologically meaningful regions and detection of cell type gradients in newly generated virtual slices.
STitch3D Tutorials and Interactive 3D Results
- Tutorials and interactive 3D results
- Example 1: adult mouse brain dataset.
- Run STitch3D on the adult mouse brain dataset
- Visualization of UMAP plots
- 3D visualization of STitch3D’s detected layer organizations of the cerebral cortex
- 3D visualization of STitch3D’s reconstructed curve-shaped distributions of four hippocampal neuron types
- 3D visualization of STitch3D’s generated virtual slice
- Example 2: human embryonic heart dataset.
- Run STitch3D on the human heart dataset
- 3D visualization of STitch3D’s identified anatomical regions
- 3D visualization of STitch3D’s inferred spatial distributions of cardiomyocytes
- 3D visualization of STitch3D’s spatiotemporal cell-type deconvolution results
- 3D visualization of gene expression patterns
- Differentially expressed gene analysis for developing human hearts
- Example 3: Drosophila embryo dataset.
- Example 4: human dorsolateral prefrontal cortex (DLPFC) dataset.
- Example 5: HER2-positive breast cancer dataset.
- Example 1: adult mouse brain dataset.