This code is just a refactoring of the original code associated with the article DeepPPI: Boosting Prediction of Protein-Protein Interactions with Deep Neural Networks. by Du et al, J Chem Inf Model. 2017 Jun 26;57(6):1499-1510. doi: 10.1021/acs.jcim.7b00028. Epub 2017 May 26. The sole purpose is to make the code run with Keras >= 2.0. All the credit for the analysis goes to the authors. The original code and data can be retrieved from this website.

##DeepPPI

Theano/Keras implementation of a deep learning network designed for Protein Protein Interaction Prediction.

###DeepPPI uses the following dependencies:

• python 2.7
• numpy 1.11.1
• scipy 0.18.1
• HDF5 and h5py
• scikit-learn 0.18
• Theano 0.8.2
• keras 1.2.0

###Guiding principles: Test_DeepPPI_Sep_Con_On_Five_Subset: DeepPPI compare with different architectures. The prediction result of our method on the five S. cerevisiae subset.

DeepPPI_Compare_With_Traditional_Algorithm_With_Optimizing: We compare DeepPPI with several traditional prediction approaches, including Nearest Neighbors, SVM, Decision Tree, Random Forest,AdaBoost, Naive Bayes, Quadratic Discriminant Analysis(QDA).

DeepPPI_Compare_Existing_Methods: Performance on the S. cerevisiae core subset, H. pylori Dataset and Human Dataset

DeepPPI_Corss_Vaild_On_Three_Dataset: Five fold cross Vaild on the S. cerevisiae core subset, H. pylori Dataset and Human Dataset

DeepPPI_Predict_Five_Across_Species: we use all 34,514 samples of the S. cerevisiae dataset as the training set and other species datasets (E. coli, C. elegans, H. sapiens, H. pylori and M.musculus) as test sets.

DeepPPI_Corss_Vaild_On_Human/Yeast_Gold_Dataset: DeepPPI uses ten fold cross validation on human/yeast gold dataset.

DeepPPI_Corss_Vaild_On_Human/Yeast_Silver_Dataset: DeepPPI uses ten fold cross validation on human/yeast silver dataset.

DeepPPI_Predict_In_All_Human/Yeast_Dataset: DeepPPI training in Gold and Silver Datasets and Testing in All Interactions Dataset

Traditional_Algorithm_Optimizing: We use grid search to find the optimal parameters of traditional algorithms

###Runing

run these file from command line.

For example:

>python DeepPPI_Predict_Five_Across_Species.pyc
output:
E.coli Acc:92.19
C.elegans Acc:94.84
H.sapiens Acc:93.77
H.pylori Acc:93.66
M.musculus Acc:91.37