TENGU (Transcript-signal ENrichment and Grouping Unit) is a Python package for processing and analyzing 10x Genomics Visium HD spatial transcriptomics data. It provides tools for aggregating 2µm bins into expression profiles of segmented units.
- Aggregation of transcription peaks to approximate cell boundaries with a cell simulation option.
- Deconvolution of bin-level data to
AnnDataorzarrobjects of segmented units. - Secondary analyses such as cell type annotation and cell-cell communication.
You can install the package directly from this GitHub repository:
pip install git+https://github.com/visanuwan/TENGU.gitFor this demonstration, we use a publicly available 10X Visium HD Mouse Brain (FFPE) dataset analyzed using Space Ranger 4.0.1. The input, output, and supplemental files for the Visium HD Gene Expression dataset can be downloaded from Visium HD Spatial Gene Expression Library, Mouse Brain (FFPE). The Slide File (H1-8CPZTMY.vlf) required by Space Ranger in this demonstration can be downloaded from this link. The predictive model used for cell type annotation can be downloaded from CellTypist's Mouse_Whole_Brain. The cell-cell communication database file can be downloaded from CellNEST database.
## Go to an example folder
cd example
## Download outputs of 10X Visium HD Mouse Brain and put them into a counts folder
## Download inputs of 10X Visium HD Mouse Brain and put them into a data folder
## Download CellTypist's Mouse_Whole_Brain and put it into a model folder
## Download CellNEST's cell-cell communication database file and put it into a database folder
## (1) Run tengu segmentation
tengu segmentation --counts counts --image data/Visium_HD_Mouse_Brain_tissue_image.tif --out_h5ad TENGU_segmented.h5ad --out_cellmark TENGU_cellmark.csv
## (2) Run Space Ranger with segmented units identified by TENGU
spaceranger count --id=TENGU_Mouse_Brain --fastqs=data/Visium_HD_Mouse_Brain_fastqs --transcriptome=/path/to/refdata-gex-GRCm39-2024-A --probe-set=/path/to/probe_sets/Visium_Mouse_Transcriptome_Probe_Set_v2.1.0_GRCm39-2024-A.csv --slide=H1-8CPZTMY --jobmode=local --area=A1 --image=data/Visium_HD_Mouse_Brain_tissue_image.tif --create-bam=true --cytaimage=data/Visium_HD_Mouse_Brain_cytaimage.tif --umi-registration=false --filter-probes=false --custom-segmentation-file=TENGU_cellmark.csv --nucleus-expansion-distance-micron=0 --output-dir=TENGU_counts
## (3) Run tengu annotation
tengu annotation --counts TENGU_counts --model model/Mouse_Whole_Brain.pkl --out predicted_celltype.csv
## (4) Run tengu communication
tengu communication --counts TENGU_counts --database database/CellNEST_database.csv --out ccc.csv
## (5) Run tengu simulation.
## The cellmark file from this step can be used with Space Ranger to generate segmented results, which can be used for TENGU's secondary analyses (annotation and communication).
tengu simulation --counts TENGU_counts --out_zarr TENGU_simulation.zarr --out_cellmark TENGU_simulation_cellmark.csv
tengu segmentation- aggregate expression profiles of 2µm bins into segmented unitstengu annotation- annotate cell types based on a supervised machine-learning classifiertengu communication- derive a basic spatially aware cell-cell communicationtengu simulation- denoise and resolve transcript diffusion with a probabilistic cell simulation framework
Please cite the following article if you use TENGU in your research
Visanu Wanchai, Nancy C. Bustamante-Gomez, Alongkorn Kurilung, Karen E. Beenken, Sergio Cortes, Mark S. Smeltzer, Yuet-Kin Leung, Jinhu Xiong, Maria Almeida, Charles A. O'Brien, Intawat Nookaew, A transcriptomic-driven segmentation and cell simulation framework for high-resolution spatial transcriptomics and cell-cell communication. bioRxiv 2026.04.24.720489; doi: https://doi.org/10.64898/2026.04.24.720489
TENGU is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
