generate_vis.md

Mode: generate_vis (manual visualization)

In most cases you do not need this mode. Running run_dataset with do_visu="true" builds the visualization automatically — it extracts the image/coordinates/scale factor from the spatial object and computes the clustering for you (see docs/run_dataset.md / SETUP.md). Use generate_vis only when you already have the proportions, image, coordinates and a clustering as separate files and want to (re)build the HTML from them.

Purpose

Builds an interactive HTML report overlaying deconvolution proportions on a Visium image and displaying spot clustering, from pre-computed inputs.

Required configuration

Key	Type	Example
mode	string	"generate_vis"
sp_input	file	"sample_sp.rds"
norm_weights_filepaths	csv list	"props_rctd.tsv,props_cell2location.tsv"
st_coords_filepath	CSV file	"tissue_positions_list.csv"
data_clustered	CSV file	"seurat_metadata.csv"
image_path	image	"tissue_hires.png"
deconv_methods	csv list	"rctd,cell2location"

Optional configuration

Key	Default	Notes
output	.	Directory where the HTML file will be written.
n_largest_cell_types	"5"	How many top cell types to show per spot.
scale_factor	1.0	Scaling factor for pixel coordinates.

Example

snakemake -s main.smk \
  --config mode="generate_vis" sp_input="UKF243_T_ST_1_raw.rds" output="vis_output" \
  norm_weights_filepaths="props_rctd.tsv,props_cell2location.tsv" \
  st_coords_filepath="tissue_positions_list_243.csv" data_clustered="seurat_metadata.csv" \
  image_path="tissue_hires.png" scale_factor=0.24414062 deconv_methods=rctd,cell2location