Preparing supporting dataset

After installation, a supporting dataset need to be prepared. This contains the base and randomized networks populated with pre-calculated protein-protein semantic similarity (semsim) for the network propagation, as well as various genesets for enrichment analysis. For details, refer to the publication.

The supporting dataset for human (ID 9606) with a full IntAct network can be directly downloaded. Alternatively, the users can create customized reference networks (e.g., for other species or networks) using a Nextflow pipeline provided separately.

Downloading the pre-made dataset

The pre-made supporting data (slim version containing only “gic” semantic similarity) is deposited on Zenodo. We recommend using the zenodo-get packages for downloading. Note that the dataset require ~15G storage.

pip install zenodo-get
cd path/to/desired/folder #change this
zenodo_get 10839901

Decompressed into a desired path. Note that the compressed file contain a top-layer folder, to ensure the integrity of folder structure after decompression. Do not modify the folder structure.

cd path/to/desired/folder #Change this.
tar -xzf support_data_slim.tar.gz

If tar is not available, the following Python code might be used:

import tarfile

target_folder = "path/to/desired/folder" #Change this.

def decompress_tar_gz(file_path, output_dir):
    with tarfile.open(file_path, "r:gz") as tar:
        tar.extractall(path=output_dir)

decompress_tar_gz("./support_data_slim.tar.gz", target_folder)

Creating dataset with customized reference network (Beta)

The Nextflow pipeline for creating customized support data is currently being tested. User that wishes to use this pipeline can contact the authors for further information.

A user-defined network will be required as input. This can be of human or other model species. The pipeline will then calculate the semantic similarity for all protein pairs, create randomized network and populate the semantic similarity. The user can also use different types of semantic similarity metric where they see suitable.

For large networks such as the IntAct used in the publication, the pipeline would require large memory and multiple cores to be executed. Thus, the pipeline should be run on a server rather than personal computers.