XGBDTI: An XGBoost Ensemble Approach for Network-Based Drug-Target Interaction Prediction

XGBDTI is a computational pipeline to predict novel drug-target interactions (DTIs) from heterogeneous network.

Requirements

The dependent packages are as following. They are easy to download using pip except xgboost。

• xgboost (You can download & install this package following(https://blog.csdn.net/mmc2015/article/details/44623301))
• numpy
• sklearn
• pandas
• scipy
• matplotlib

Quick Start

You can just run python2 main.py to reproduce the cross validation results of XGBDTI. Options are:

• -r: negtive positive ratio, one or ten

Tutorial

Of course, you can go through all the process by run python2 totalPipeline.py Options are:

• -d: drug network list, seperated by comma, without suffix.
• -p: protein network list, seperated by comma, without suffix.
• -dm: drug feature vector dimensions
• -pm: protein feature vector dimensions
• -r: negtive positive ratio As for more hyperparameters in XGBoost model, you can modify them in the src/model.py

Code & Data

src/ directory

.
├── main.py
├── utils.py
├── totalPipeline.py
└── tuning_xgboost.py

• utils.py: Contain all the important methods and functions in this project
• main.py: XGBDTI quick start
• tuning_xgboost.py: Greedy grid search for XGBoost hyperparameters
• totalPipeline.py: Go through all the process for XGBDTI

Data/ directory

../Data/
├── InteractionData
│   ├── disease.txt
│   ├── drug.txt
│   ├── mat_drug_disease.txt
│   ├── mat_drug_drug.txt
│   ├── mat_drug_protein.txt
│   ├── mat_drug_protein_disease.txt
│   ├── mat_drug_protein_drug.txt
│   ├── mat_drug_protein_homo_protein_drug.txt
│   ├── mat_drug_protein_sideeffect.txt
│   ├── mat_drug_protein_unique.txt
│   ├── mat_drug_se.txt
│   ├── mat_protein_disease.txt
│   ├── mat_protein_drug.txt
│   ├── mat_protein_protein.txt
│   ├── protein.txt
│   └── se.txt
├── SimilarityData
│   ├── Similarity_Matrix_Drugs.txt
│   └── Similarity_Matrix_Proteins.txt
├── drug_dict_map.txt
└── protein_dict_map.txt

• Similarity_Matrix_Drugs.txt : Drug & compound similarity scores based on chemical structures of drugs ([0,708) are drugs, the rest are compounds).
• Similarity_Matrix_Proteins.txt : Protein similarity scores based on primary sequences of proteins.
• mat_drug_protein_homo_protein_drug.txt : Drug-Protein interaction matrix, in which DTIs with similar drugs (i.e., drug chemical structure similarities > 0.6) or similar proteins (i.e., protein sequence similarities > 40%) were removed.
• mat_drug_protein_drug.txt: Drug-Protein interaction matrix, in which DTIs with drugs sharing similar drug interactions (i.e., Jaccard similarities > 0.6) were removed.
• mat_drug_protein_sideeffect.txt: Drug-Protein interaction matrix, in which DTIs with drugs sharing similar side effects (i.e., Jaccard similarities > 0.6) were removed.
• mat_drug_protein_disease.txt: Drug-Protein interaction matrix, in which DTIs with drugs or proteins sharing similar diseases (i.e., Jaccard similarities > 0.6) were removed.
• mat_drug_protein_unique: Drug-Protein interaction matrix, in which known unique and non-unique DTIs were labelled as 3 and 1, respectively, the corresponding unknown ones were labelled as 2 and 0.

The above files are used to test the robustness of the model (see https://github.com/FangpingWan/NeoDTI). In addition to them, other files are provided by the project.

DataSet/ directory

pre-trained data set.
The integer denotes the ratio of negative to positive.

feature/ directory

pre-trained vector representations for drugs and proteins.
The integer denotes the dimension of the feature vectors.

network/ directory

pre-computed similarity matrices.

whole/ directory

This directory stores the data and program for test on the whole drug-target interaction network. And the drugs(or proteins) with no interaction was moved.
Due to time pressure, I did not add comments. So the script in this folder is a bit confusing.

.
├── ConstructModel.py
├── big.model
├── constructData.py
├── drug50Refined.txt
├── mat_p_d_Refined.txt
├── predict.py
├── protein200Refined.txt
└── totalX.txt

• constructData.py construct data set
• ConstructModel.py training model
• predict.py make predictions
• big.model pre-trained model file
• mat_p_d_Refined.txt protein-drug interaction matrix remove all-zero lines and columns.
• drug50Refined.txt drug features removing non-interaction items
• protein200Refined.txt protein features removing non-interaction items.
• totalX.txt combined data using Cartesian product

supplement/ directory