dtgz-sudo/npi-hgnn

NPI-HGNN

The proposed NPI-HGNN is an end-to-end graph neural network (GNN)-based approach for the identification of NPIs from a large heterogeneous network, consisting of the ncRNA-protein interaction network, the ncRNA-ncRNA similarity network, and the protein-protein interaction network. NPI-HGNN consists of five modules, including the construction of reliable negative samples, extraction of enclosing subgraphs, feature encoding of nodes, graph representation learning, and prediction.

Note: The Python packages and their versions needed in our source code have been provided in the ‘requirements.txt’ file on our GitHub page.
Note: NPI-HGNN does not rely on other third-party software/protocol, and it can run on any computational platforms/personal computer with CPU/GPU processors. Storage needed in NPI-HGNN depend on the size of dataset handled in experiments. 

1. Running process

1.1 Generating kmer ncRNA-ncRNA similarity network

Python .\src\npi_hgnn\generate_rrsn.py --dataset {datasetName} --ratio {ratio}

Parameters

• --dataset : Dataset name. [NPHN3265 | NPHN4158 | NPHN7317 | NPHN-Homo | NPHN-Mus]
• --ratio : The ration of edges generated in the ncRNA-ncRNA similarity network to min(NPIs,PPIs)

1.2 Generating kmer protein-protein similarity matrix

Python .\src\npi_hgnn\generate_ppsm.py --dataset {datasetName} --ratio {ratio}

Parameters

• --dataset : Dataset name. [NPHN3265 | NPHN4158 | NPHN7317 | NPHN-Homo | NPHN-Mus]

1.3 Negative sample selection and partitioning the dataset

Python .\src\npi_hgnn\generate_edgelist.py --dataset {datasetName} --samplingType {no} --num_fold {num}

Parameters

• --dataset : Dataset name. [NPHN3265 | NPHN4158 | NPHN7317 | NPHN-Homo | NPHN-Mus]
• --samplingType : [0 | 1 | 2 | 3].
• --num_fold : [integer greater than 0]

1.4 Extracting kmer substrings of ncRNA and protein sequences

Python .\src\npi_hgnn\generate_kmer.py --dataset {datasetName} --kRna {num} --kProtein {num}

Parameters

• --dataset : Dataset name. [NPHN3265 | NPHN4158 | NPHN7317 | NPHN-Homo | NPHN-Mus]
• --kRna : [3].
• --kProtein : [2]

1.5 Generating kmer occurrence vector

Python .\src\npi_hgnn\generate_frequency.py --dataset {datasetName}

Parameters

• --dataset : Dataset name. [NPHN3265 | NPHN4158 | NPHN7317 | NPHN-Homo | NPHN-Mus]

1.6 Generating pyfeat vector

Python .\src\npi_hgnn\generate_pyfeat.py --dataset {datasetName}

Parameters

• --dataset : Dataset name. [NPHN3265 | NPHN4158 | NPHN7317 | NPHN-Homo | NPHN-Mus]
• --other_parameters : please see https://github.com/mrzResearchArena/PyFeat

1.7 Generating node feature vector

Python .\src\npi_hgnn\generate_node_vec.py --dataset {datasetName} --nodeVecType {no} --samplingType {no} --subgraph_type {no} --fold {no}

Parameters

• --dataset : Dataset name. [NPHN3265 | NPHN4158 | NPHN7317 | NPHN-Homo | NPHN-Mus]
• --nodeVecType : node feature type. [0 | 1 | 2].
• --samplingType : [0 | 1 | 2 | 3].
• --subgraph_type : [0 | 1 | 2].
• --fold : which fold is this. [0 | 1 | 2 | 3 | 4].
• --other_parameters : please see https://github.com/aditya-grover/node2vec

1.8 Generating dataset

Python .\src\npi_hgnn\generate_dataset.py --dataset {datasetName} --nodeVecType {no} --samplingType {no} --subgraph_type {no} --fold {no}

Parameters

• --dataset : Dataset name. [NPHN3265 | NPHN4158 | NPHN7317 | NPHN-Homo | NPHN-Mus]
• --nodeVecType : node feature type. [0 | 1 | 2].
• --samplingType : [0 | 1 | 2 | 3].
• --subgraph_type : [0 | 1 | 2].
• --fold : which fold is this. [0 | 1 | 2 | 3 | 4].

1.9 Training

Python .\src\npi_hgnn\train.py --dataset {datasetName} --nodeVecType {no} --samplingType {no} --subgraph_type {no} --fold {no}

Parameters

• --dataset : Dataset name. [NPHN3265 | NPHN4158 | NPHN7317 | NPHN-Homo | NPHN-Mus]
• --nodeVecType : node feature type. [0 | 1 | 2].
• --samplingType : [0 | 1 | 2 | 3].
• --subgraph_type : [0 | 1 | 2].
• --fold : which fold is this. [0 | 1 | 2 | 3 | 4].
• --other_parameters : default value.

2. Case study

2.1 Generating case study dataset

Python .\src\case_study\case_study_dataset.py

Parameters

• --all_parameters : default value.

2.2 Case study training

Python .\src\case_study\case_study_train.py

Parameters

• --all_parameters : default value.

2.3 Case study prediction

Python .\src\case_study\case_study_predict.py --epoch {no}

Parameters

• --epoch : [integer greater than 0].
• --other_parameters : default value.