scPML: Pathway-based Multi-view Learning for Cell Type Annotation from Single-cell RNA-seq Data

A PyTorch implementation of scPML.

Requirements

• R >= 4.2.1
• python >= 3.8.12

Python package version

• scikit-learn >= 1.0.2
• numpy >= 1.23.3
• pytorch >= 1.10.2
• torch-geometric >= 2.0.3
• networkx >= 2.8.4
• pandas >= 1.4.3
• scipy >= 1.9.1 We recommend upgrading all packages to the latest version.

Input Data

When using your own data, you have to provide:

• the raw data matrix of training data and cells labels.
• the raw data matrix of test data. Training and test data should share the same gene features.

For the sake of convenience, we use a directory tree to organize experiments and data:

-- proj_name
  -- raw_data
    -- ref
        data_1.csv
        label_1.csv
    -- query
        data_1.csv
  -- data
    -- ref
    -- query       

When using your own data, put raw training data matrix in the raw_data/ref/ and name it data_1.csv, raw training label in raw_data/ref/ and name it label_1.csv. Put the raw test data in the raw_data/query/ directory.

The data directory contains the pre-processed data. More details can be seen in the demo directory.

Run the demo

Similarity matrix construction

    cd demo
    Rscript ..\utils\get_sm.R seq_well_10x_v3    

Data preprocess

    Rscript ..\utils\pre_process.R seq_well_10x_v3
    python ..\utils\data_csv2h5.py --path=seq_well_10x_v3 --subpath=raw_data
    python ..\utils\data_csv2h5.py --path=seq_well_10x_v3 --subpath=data    

Run scPML

    python main.py

Output

The results will be stored in the result folder.

Reproducibility

We provide hyper-parameters for all experiements with a pickle file named hyper_parameters.pkl. The data is a dict. You can load the file by using pickle.

import pickle as pkl

with open('hyper_paramters.pkl', 'rb') as f:
    params = pkl.load(f)