SMINT: Spatial Multi-Omics Integration¶
SMINT is a Python package for Spatial Multi-Omics Integration with enhanced segmentation capabilities and streamlined workflow.
Overview¶
SMINT provides a comprehensive toolkit for processing and analyzing spatial omics data, including:
- Multi-GPU cell segmentation for whole-slide images
- Distributed segmentation using Dask for improved performance
- Live segmentation monitoring with intuitive visualization tools
- Registration of spatial transcriptomics with metabolomics and post-stain imaging
- Integration with R analysis scripts
- Comprehensive documentation with step-by-step guides
- HPC deployment scripts for SLURM-based clusters
Key Features¶
Enhanced Segmentation¶
- Multi-GPU Support: Utilize multiple GPUs for faster processing of large whole-slide images
- Distributed Computing: Use Dask to distribute segmentation tasks across multiple nodes
- Live Monitoring: Track segmentation progress in real-time with the built-in viewer
- Adaptive Segmentation: Automatically adjust segmentation parameters for optimal results
- Dual-Model Segmentation: Simultaneously segment cells and nuclei with specialized models
Registration¶
- Two Regimes: STalign LDDMM for sequential sections; correspondence fitting (RANSAC / affine / TPS) for post-staining on the same section
- Coarse Pre-registration: Scale to a reference coordinate system, rotate and flip before fine registration
- Manual Landmarks: Interactive landmark annotation, in napari or standalone
- Honest Metrics: Target Registration Error on held-out pairs, so a transform cannot flatter itself by memorising correspondences
- HPC Native: Submit to SLURM (CPU or GPU) or run locally, from Python or napari
R Integration¶
- Seamless Python-R Bridge: Call R scripts and functions directly from Python
- Data Transfer: Convert data between Python and R formats
- Existing R Scripts: Use your existing R analysis scripts within the SMINT workflow
Visualization¶
- Live Viewer: Monitor segmentation progress with a live viewer
- Segmentation Overlays: Visualize segmentation results overlaid on the original image
- Feature Plots: Generate feature plots and spatial heatmaps
HPC Deployment¶
- SLURM Integration: Ready-to-use SLURM submission scripts for HPC deployment
- Resource Management: Optimized resource allocation for different processing stages
- Checkpointing: Resume processing from checkpoints after interruptions
Getting Started¶
- Installation: Install SMINT and its dependencies
- Segmentation: Run cell segmentation on whole-slide images
- Alignment: Align spatial transcriptomics data
- R Integration: Use R scripts and functions with SMINT
- Examples: Complete examples of SMINT workflows
- Configuration: Configure SMINT for your specific needs
- HPC Deployment: Run SMINT on HPC clusters
- Visualizations: Interactive 3D multi-omics plots from the BCRL patient cohort
Citation¶
If you use SMINT in your research, please cite: