SONAR: A Probabilistic Framework for Cell-Type Deconvolution in Spatial Transcriptomics

SONAR is an algorithm developed for cell-type deconvolution in spatial transcriptomics. It integrates spatial information in a balanced way to enhance performance and robustness.

Key Features

• Enhanced Signal-to-Noise Ratio
  Utilizes the similarity of spatial locations to boost the signal-to-noise ratio, helping to extract meaningful patterns from the data.

• Local Spatial Heterogeneity Consideration
  Incorporates local spatial heterogeneity to avoid over-reliance on spatial information, ensuring that the inherent biological diversity is accurately captured.

• Robust Probabilistic Framework
  Built on a probabilistic framework, SONAR delivers more robust and reliable results compared to deep learning–based approaches.

Installation

#library(devtools)

devtools:: install_github("lzygenomics/SONAR")

Dependence

• R version >= 4.0.5.
• R packages:
  • this.path>=0.5.1; Matrix>=1.3.4; data.table>=1.14.0; Seurat>=4.0.3; matlabr=1.5.2; R.matlab=3.7.0
• MATLAB version >= R2019a
  • Just install the MATLAB, subsequent operations do not require you to use the MATLAB language

Run SONAR

1. Install the dependence (pay attention that you need to install the MATLAB)

2. Install SONAR

3. Download the SONAR files (This files structure will help you run on the custom data)

4. Open Example/SONAR-entrance.Rmd , you could run and get the example results.

For running the custom dataset, you could modify the data preparation stage in Example/SONAR-entrance.Rmd with the same format, and substitude the single cell data and spatial data in this files structure

Files Annotation

• /inst/extdata/

  These files store the required input information, See SONAR-entrance.Rmd for specific format requirements.

• /result/

  These files store the output results.

  Proportions of all cell types in each spot (SONAR.results.txt);

  Spatial pie plots (pie.pdf);

  Spatial distribution of specific cell types proportion (abs_prop.pdf / scaled_prop.pdf);

  Colocalization(correlation along the spatial) for pairs of cell types (colocalization.pdf).

• /core-code/

  no need to operate.

  These files store the core code, and store the preprocessed data that delivered to SONAR.

A brif Example

Please Follow the Example/SONAR-entrance.Rmd

Publication Link

https://www.nature.com/articles/s41467-023-40458-9

How to cite SONAR

Liu, Z., Wu, D., Zhai, W. et al. SONAR enables cell type deconvolution with spatially weighted Poisson-Gamma model for spatial transcriptomics. Nat Commun 14, 4727 (2023). https://doi.org/10.1038/s41467-023-40458-9

Thank you, and happy researching!