10X Xenium Starter Tutorial

Input Data

This tutorial uses SGE data generated with the 10x Genomics Xenium platform, and it has been cropped to a small region of the adult mouse brain for demonstration purposes.

File Format

10X Genomics Xenium platform outputs Spatial Gene Expression (SGE) data in a comma-separated values (CSV) format.

CSV File Format

"transcript_id","cell_id","overlaps_nucleus","feature_name","x_location","y_location","z_location","qv"
281827164036151,133793,1,"Sox10",2350.0232,4153.6846,16.592316,40.0
281827164036152,133793,1,"Sox10",2350.2585,4154.5225,17.237207,10.514394
281827164036164,151216,0,"Sox10",2350.5874,4277.699,14.285685,40.0

• "transcript_id": Unique identifier for each detected transcript molecule.
• "cell_id": ID of the segmented cell associated with the transcript.
• "overlaps_nucleus": 1 if the transcript overlaps the nucleus mask, 0 otherwise.
• "feature_name": Gene or other features name corresponding to the transcript.
• "x_location": X-coordinate of the transcript.
• "y_location": Y-coordinate of the transcript.
• "z_location": Z-coordinate (depth) of the transcript.
• "qv": Quality value indicating confidence in transcript detection.

Data Access

The example data is hosted on Zenedo ().

Follow the commands below to download the example data.

work_dir=/path/to/work/directory
cd $work_dir
wget  https://zenodo.org/records/15701394/files/xenium_starter.raw.tar.gz 
tar --strip-components=1 -zxvf xenium_starter.raw.tar.gz  

Set Up the Environment

Define paths to all required binaries and resources, and target AWS S3 bucket. Optionally, specify a fixed color map for consistent rendering.

# ====
# Replace each placeholder with the actual path on your system.  
# ====
work_dir=/path/to/work/directory        # path to work directory that contains the downloaded input data
cd $work_dir

# Define paths to required binaries and resources
spatula=/path/to/spatula/binary         # path to spatula executable
punkst=/path/to/punkst/binary           # path to FICTURE2/punkst executable
tippecanoe=/path/to/tippecanoe/binary   # path to tippecanoe executable
pmtiles=/path/to/pmtiles/binary         # path to pmtiles executable
aws=/path/to/aws/cli/binary             # path to AWS CLI binary

# (Optional) Define path to color map. 
cmap=/path/to/color/map                 # Path to the fixed color map for rendering. cartloader provides a fixed color map at cartloader/assets/fixed_color_map_256.tsv.

# AWS S3 target location for cartostore
AWS_BUCKET="EXAMPLE_AWS_BUCKET"         # replace EXAMPLE_AWS_BUCKET with your actual S3 bucket name

# Activate the bioconda environment
conda activate BIOENV_NAME              # replace BIOENV_NAME with your bioconda environment name

Define data ID and analysis parameters:

# Unique identifier for your dataset
DATA_ID="xenium_hippo"                  # change this to reflect your dataset name
PLATFORM="10x_xenium"                   # platform information
SCALE=1                                 # coordinate to micrometer scaling factor

# LDA parameters
train_width=12                           # define LDA training hexagon width (comma-separated if multiple widths are applied)
n_factor=6,12                            # define number of factors in LDA training (comma-separated if multiple n-factor are applied)

How to Define Scaling Factors for Xenium

The Xenium example data currently used here provides SGE in micrometer units. Use define scaling factor from coordinate to micrometer as 1.

SGE Format Conversion

Convert the raw input to the unified SGE format. See more details in Reference page.

cartloader sge_convert \
  --makefn sge_convert.mk \
  --platform ${PLATFORM} \
  --in-csv ./input.tsv.gz \
  --units-per-um ${SCALE} \
  --out-dir ./sge \
  --exclude-feature-regex '^(BLANK|NegCon|NegPrb)' \
  --sge-visual \
  --spatula ${spatula} \
  --n-jobs 10

Parameter	Required	Type	Description
--platform	required	string	Platform (options: "10x_visium_hd", "seqscope", "10x_xenium", "bgi_stereoseq", "cosmx_smi", "vizgen_merscope", "pixel_seq", "generic")
--in-csv	required	string	Path to the input TSV/CSV file
--units-per-um	required	float	Scale to convert coordinates to microns (default: 1.0)
--out-dir	required	string	Output directory for the converted SGE files
--makefn		string	File name for the generated Makefile (default: sge_convert.mk)
--exclude-feature-regex		regex	Pattern to exclude control features
--sge-visual		flag	Enable SGE visualization step (generates diagnostic image) (default: FALSE)
--spatula		string	Path to the spatula binary (default: spatula)
--n-jobs		int	Number of parallel jobs for processing (default: 1)

FICTURE analysis

Compute spatial factors using punkst (FICTURE2 mode). See more details in Reference page.

cartloader run_ficture2 \
  --makefn run_ficture2.mk \
  --main \
  --in-transcript ./sge/transcripts.unsorted.tsv.gz \
  --in-feature ./sge/feature.clean.tsv.gz \
  --in-minmax ./sge/coordinate_minmax.tsv \
  --cmap-file ${cmap} \
  --exclude-feature-regex '^(mt-.*$|Gm\d+$)' \
  --out-dir ./ficture2 \
  --width ${train_width} \
  --n-factor ${n_factor} \
  --spatula ${spatula} \
  --ficture2 ${punkst} \
  --n-jobs 10 \
  --threads 10

Parameter	Required	Type	Description
--main	required 1	flag	Enable cartloader to run all five steps
--in-transcript	required	string	Path to input transcript-level SGE file
--out-dir	required	string	Path to output directory
--width	required	int or comma-separated list	LDA training hexagon width(s)
--n-factor	required	int or comma-separated list	Number of LDA factors
--makefn		string	File name for the generated Makefile (default: run_ficture2.mk )
--in-feature		string	Path to input feature file
--in-minmax		string	Path to input coordinate min/max file
--cmap-file		string	Path to color map file
--exclude-feature-regex		regex	Pattern to exclude features
--spatula		string	Path to the spatula binary (default: spatula)
--ficture2		string	Path to the punkst directory (defaults to punkst repository within submodules directory of cartloader)
--n-jobs		int	Number of parallel jobs (default: 1)
--threads		int	Number of threads per job (default: 1)

_{¹: cartloader requires the user to specify at least one action. Available actions includes: --tile to run tiling step; --segment to run segmentation step; --init-lda to run LDA training step; --decode to run decoding step; --summary to run summarization step; --main to run all above five actions.}

cartloader Compilation

Generate pmtiles and web-compatible tile directories. See more details in Reference page.

cartloader run_cartload2 \
  --makefn run_cartload2.mk \
  --fic-dir ./ficture2 \
  --out-dir ./cartload2 \
  --id ${DATA_ID} \
  --spatula ${spatula} \
  --pmtiles ${pmtiles} \
  --tippecanoe ${tippecanoe} \
  --n-jobs 10 \
  --threads 10

Parameter	Required	Type	Description
--fic-dir	required	string	Path to the input directory containing FICTURE2 output
--out-dir	required	string	Path to the output directory for PMTiles and web tiles
--id	required	string	Dataset ID used for naming outputs and metadata
--makefn		string	File name for the generated Makefile (default: run_cartload2.mk)
--spatula		string	Path to the spatula binary (default: spatula)
--pmtiles		string	Path to the pmtiles binary (default: pmtiles)
--tippecanoe		string	Path to the tippecanoe binary (default: tippecanoe)
--n-jobs		int	Number of parallel jobs (default: 1)
--threads		int	Number of threads per job (default: 1)

Upload to Data Repository

AWS Uploads

Copy the generated cartloader outputs to your designated AWS S3 catalog path:

cartloader upload_aws \
  --in-dir ./cartload2 \
  --s3-dir "s3://${AWS_BUCKET}/${DATA_ID}" \
  --aws ${aws} \
  --n-jobs 10

Parameter	Required	Type	Description
--in-dir	required	string	Path to the input directory containing the cartloader compilation output
--s3-dir	required	string	Path to the target S3 directory for uploading
--aws		string	Path to the AWS CLI binary
--n-jobs		int	Number of parallel jobs