mjuailab/sccap repository overview

⁠scCap: Single-cell Knowledge-augmented Clustering for Annotation-free Phenotype Prediction

scCap is an annotation-free framework that constructs clusters through knowledge-augmented clustering and leverages them to enable accurate and interpretable phenotype prediction.

This Docker image includes a fully configured environment for:

• Preprocessing scRNA-seq datasets

• Encoding cells with pretrained single-cell foundation models

• Knowledge-augmented clustering (Initialization → Refinement)

• Hierarchical MIL–based phenotype prediction with dual-level attention

⁠Quickstart

# Step 1: Clone the repository
git clone https://github.com/mjuailab/scCap.git
cd scCap

# Step 2: Pull the prebuilt Docker image
docker pull mjuailab/sccap:latest

# Step 3: Run the container with GPU support and sufficient shared memory
docker run -it --gpus all --shm-size=[shared_memory_size] \
  -v [local_project_directory]:/workspace \
  mjuailab/sccap:latest

⁠Documentation

Full documentation and tutorials are available on GitHub⁠