ST-CNABench

ST-CNABench is a public benchmark framework for copy number alteration inference on spatial transcriptomics. It provides one controller for dataset preparation, model execution, and evaluation through a unified prep -> run -> eval workflow.

The current public release includes 9 CNA inference methods.

Included Methods

• CalicoST
• CopyKAT
• InferCNV
• Clonalscope (Clonalscope_NoWGS, Clonalscope_WGS)
• Numbat
• Xclone
• SCEVAN
• STARCH
• SlideCNA

Installation

git clone https://github.com/YangLabHKUST/ST-CNABench.git
cd ST-CNABench
conda create -n benchmark_env python=3.10 -y
conda activate benchmark_env
pip install -e .
st-cnabench --help

For method-specific runtime environments, external tools, and reference data, see the installation guide: https://cnadocs.readthedocs.io/en/latest/installation/

Run

Prepare data:

st-cnabench --steps prep \
  --data-config <DATA_CONFIG> \
  --prep-ids <DATASET_ID>

Run methods:

st-cnabench --steps run \
  --data-config <DATA_CONFIG> \
  --model-config <MODEL_CONFIG> \
  --prep-ids <DATASET_ID> \
  --exec-mode <conda|docker|apptainer> \
  --models <METHOD_1> <METHOD_2>

Evaluate results:

st-cnabench --steps eval \
  --data-config <DATA_CONFIG> \
  --eval-config <EVAL_CONFIG> \
  --prep-ids <DATASET_ID> \
  --models <METHOD_1> <METHOD_2> \
  --eval-tasks <TASK_NAME>

Tutorials

We provide demo data and tutorial workflows for:

• the packaged cSCC demo pipeline
• cna_profile task example
• tumor_normal task example
• subclone identification task example

Documentation site: https://cnadocs.readthedocs.io/en/latest/

Quickstart tutorial: https://cnadocs.readthedocs.io/en/latest/tutorials/quickstart_demo/

Contact

HAN Shi
shanav@connect.ust.hk