INDTI: Drug target interactions prediction based on interactive inference network

Overview of INDTI

^{INDTI has an embedding layer, an encoding layer, an interaction layer, a feature extraction layer, and an output layer. At the embedding layer, drug sequences and target molecules are embedded. The embedding of drugs and targets is encoded at the encoding layer. The interaction layer simulates drug-target interactions. In the feature extraction layer, the interaction features of the interaction matrix are extracted and the prediction results are obtained. In general, when predicting the interaction between a target and a drug, most models connect the extracted features of both the drug and the target molecular sequences, feeding them into the prediction model. However, our model uses the dot product to generate a scalar indicating the interaction strength between a single target-drug minimum unit pair, which produces interpretable model interaction predictions.}

Files

train.py: Training function of INDTI:
Set the data set and train the model
First, modify the corresponding configuration in the config.py file. Then select the dataset for training
test.py: Test function of INDTI:
Specify the data to be used for prediction, make model prediction, and generate interactive files at the same time
First, modify the corresponding configuration in the config.py file. Then select the test set, and refer to val.csv for the format of the test set
stream.py: Used to generate embedded
models.py: INDTI model
config.py: The config of INDTI:
Type1 represents Direct embedding + CNN
Type2 represents Direct embedding + Molecular fingerprint + CNN
Type3 represents Subsequent embedding + Attention
Type4 represents Subsequent embedding + Molecular fingerprint + Attention
fp_type is used to specify PubChem or Morgan
CNN.py: The CNN encoder
self_attention.py: The Self-Attention encoder
chord: Example of interactive file
data: Example of preprocessing data
dataset: Data from three datasets: BindingDB,BIOSNAP and DAVIS
ESPF: Generate embedded related files
model: Example of Trained model(https://stumail-my.sharepoint.cn/:u:/g/personal/21yqchen2_stu_edu_cn/Edr1wyTOvAFEnLRpf6Sjqr4BKxqUr0KoAgD67lYuCc27jA)