scDesign3 generates realistic in silico data for multimodal single-cell and spatial omics.

Area of Science:

• Computational Biology
• Single-cell Omics
• Spatial Omics

Background:

• Generating realistic single-cell and spatial omics data is crucial for developing and benchmarking computational tools.
• Existing methods may lack the flexibility to capture the complexity of diverse omics data types and experimental designs.

Purpose of the Study:

• To introduce scDesign3, a statistical simulator for generating high-fidelity single-cell and spatial omics data.
• To provide a unified probabilistic framework for modeling both single-cell and spatial omics data.
• To enable the generation of in silico controls for robust computational tool evaluation.

Main Methods:

• Developed scDesign3, a statistical simulator that learns interpretable parameters from real omics data.
• Employed a unified probabilistic model to handle various cell states, experimental designs, and feature modalities.
• Implemented methods to assess the goodness-of-fit for inferred cell clusters, trajectories, and spatial locations.

Main Results:

• scDesign3 successfully generates realistic single-cell and spatial omics data across diverse biological contexts.
• The simulator infers biologically meaningful parameters, offering insights into data generation processes.
• Generated in silico negative and positive controls facilitate effective benchmarking of computational tools.

Conclusions:

• scDesign3 provides a powerful and flexible platform for simulating complex omics data.
• The tool aids in the development and validation of computational methods for single-cell and spatial omics analysis.
• scDesign3 enhances the reliability and reproducibility of omics data analysis through in silico controls.