Support Vector Machine on Graph kernel for protein-protein conformation ranking

Minimal information to install the module

• Check if command mpiexec is available or not in your console. If not, download and install openmpi or mpich.
• Install iScore using pip install iScore

iScore offers simple solutions to classify protein-protein interfaces using a support vector machine approach on graph kernels. The simplest way to use iScore is through dedicated binaries that hide the complexity of the approach and allows access to the code with simple command line interfaces. The two binaries are iscore.train and iscore.predict that respectively train a model using a trainging set and use this model to predict the near-native character of unkown conformations.

To use these binaries easily the following file structure is advised :

root/
 |__train/
 |    |__ pdb/
 |    |__ pssm/
 |    |__ caseID.lst
 |__predict/
      |__pdb/
      |__pssm/
      |__ caseID.lst (optional)

The train subdirectory contains the PDB files and the PSSM files of the conformation contained in the training set. The PSSM files can be calculated using PSSMGen https://github.com/DeepRank/PSSMGen. To train the model simply go to the train subdirectory and type:

mpiexec -n ${NPROC} iScore.train

This binary will generate a archive file called by default training_set.tar.gz that contains all the information needed to predict binary classes of a test set using the trained model. To use this model go into the test subdirectory and type:

mpiexec -n ${NPROC} iScore.predict --archive ../train/training_set.tar.gz

This binary will output the binary class and decision value of the conformations in the test set in a text file iScorePredict.txt.