The MNHGHTA is designed for the prediction of herb-target associations. Two kernels are constructed for herbs and for targets, respectively. They are fused into one unified herb kernel and target kernel, thereby building the herb and target similar networks. Then, the hybrid graph neural network and convolution neural network combiner are applied to above similar networks and herb-target association network for accessing high-level herb and target representations. Finally, the inner product is employed to generate the association score matrix.
python 3.9
torch 1.13
torch_geometric 2.3.1
pandas 1.4.2
numpy 1.21.0
scipy 1.8.0
git clone https://github.com/not4yazwz/MNHGHTA.git
cd MNHGHTA
python main.py
main
This file contains the main script used to start the entire process of training and testing the model.
model
This file includes the definition of the model and the codes for its neural network architecture.
param
This file contains the parameter configuration.
util
This file contains a series of helper functions.
fold_1 to fold_5
These folders store the experimental data. Each folder contains the similarity matrices constructed by Gaussian kernel and cosine similarity matrix for herb and target under the current independent experiment (stored with .npz format), training sets of positive and negative samples, and testing sets of positive and negative samples.
data
This folder stores the raw data and constructed data structures.