A large-scale method to predict protein-protein interfaces based on sequence and structure information


Protein-protein interactions (PPIs) are essential to all biological processes. They represent increasingly important therapeutic targets. JET2 is a new method for accurately predicting protein-protein interfaces, understanding their properties, origins and binding to multiple partners. JET2 uses three sequence- and structure-based descriptors of protein residues: evolutionary conservation, physico-chemical properties and local geometry. Appropriate combination of these descriptors yields very precise predictions for a wide range of protein-protein interfaces and discriminates them from small-molecule binding sites. Beyond its predictive power, the approach permits to dissect the interaction surfaces and unravel their complexity. The analysis of the predicted patches can foster new strategies for PPIs modulation and interaction surface redesign. JET2 is based on the Joint Evolutionary Trees (JET) method. It provides a fully automated predictive pipeline for large-scale application and allows the user to tune all paramaters depending on the biological question asked.


The JET2 package is available here.

System requirements

Linux or Mac OS X.

The program JET2 requires some external tools that should be installed:

  1. - java6

  2. -Muscle (for example v3.8), a tool for performing multiple alignment of nucleic acid and protein sequences. It is run with either Blosum62, Gonnet or HSDM matrix by automatic selection

  3. Edgar RC. (2004) Nucleic Acids Res., 32:1792-1797

  4. -Naccess v2.1.1, a program that calculates the accessible area of a molecule from a PDB (Protein Data Bank) format file

  5. Hubbard S. J., Thornton J. M. (1993), University College London

  6. -PSI-BLAST from BLAST+ Toolkit (v2.2.27 or more recent). The program can be called on the sever or locally. To reach similar outputs on the server and on a local machine, a local call to PSI-BLAST is coded in JET2, that uses the option -t 2 setting the composition-based score adjustment method conditioned on sequence properties.

  7. Altschul SF, Gish W, Miller W, Myers EW, Lipman D. (1990) J Mol Biol., 215, 403–410


  1. 1. Unzip the in the directory of your choice.

  2. 2. Set up JET2 home directory in your .bashrc or .cshrc file:

Instructions for running JET2

The instructions for running JET2 and reading the output files can be found here.

The command line to execute JET2 is:

JET2 offers a number of functionalities whose choice is controlled by the option -p:

  1. - A Compute the accessibility surface areas of the atoms and residues using Naccess

  2. - I Determine the interface residues if the PDB input file is a complex

  3. - V Compute the circular variances of the residues

  4. - J Launch Joint Evolutionary Trees analysis to evaluate residues conservation levels

  5. - C Run the clustering algorithm to define binding patches

  6. - G Insert JET2 results in the B-factors column of the input PDB file

  7. - R Evaluate JET2 results according to the real interface residues (I analysis needed)

JET2 proposes different clustering strategies whose choice is controlled by the option -a:

  1. - 0 The most appropriate strategy is automatically determined by JET2

  2. - 1 Conservation levels are used for cluster seed detection and extension

  3. - 2 Conservation levels and physico-chemical properties are used or cluster seed detection and extension

  4. - 3 SC1: the cluster seed is detected based on conservation levels; the seed is extended by mixing conservation levels and physico-chemical properties; an outer layer is added based on physico-chemical properties and circular variance (local geometry)

  5. - 4 SC2: cluster seed and extension are issued from a combination of conservation levels and circular variance; an outer layer is added based on physico-chemical properties and circular variance

  6. - 5 SC3: cluster seed, extension and outer layer are issued from a combination of physico-chemical properties and circular variance

Example of a JET2 analysis:

  1. 1. Create a working directory and place your input PDB file in it

  2. 2. Copy the default.conf configuration file from your $JET2_PATH directory and modify the locations of the software and the parameters values you wish

  3. 3. Launch JET2

       The arguments given are :

  1. - the configuration file to be used (-c)

  2. - the input PDB file (-i)

  3. - the directory where output files should be stored (-o)

  4. - the type(s) of analysis to run (-p)

  5. - the mode for PSI-BLAST call (-r)

  6. - the clustering strategy (-a)

  7. - the way the input PDB file should be treated (-d), either as a complex (complex) or by considering each chain individually (chain

The retrieval of homologous sequences from BLAST database may take some time. It is possible to do it once for all, and then run only JET2 and clustering analyses by specifying the -r input option and the location of the PSI-BLAST alignment file (-b):

An iterative version of JET2 (iJET2) is also proposed for large-scale predictions. iJET2 provides a list of consensus residues belonging to interaction patches and enables to explore the set of potentially interacting residues by varying the consensus threshold during iterations. Typically consensus residues detected in at least 2 iterations are considered as robust prediction. iJET2 is called using the -n option:

java -cp $JET2_PATH:$JET2_PATH/jet/extLibs/vecmath.jar jet.JET

    export JET2_PATH = path_of_JET2_HOME_directory (bash syntax)

    setenv JET2_PATH path_of_JET2_HOME_directory (csh syntax)

java -cp $JET2_PATH:$JET2_PATH/jet/extLibs/vecmath.jar jet.JET -c <default.conf> -i <struct.pdb> -o `pwd` -p AVJC -r local -a 3 -d chain

java -cp $JET2_PATH:$JET2_PATH/jet/extLibs/vecmath.jar jet.JET -c <default.conf> -i <struct.pdb> -o `pwd` -p AVJC -r input -b <dir_psiblast_files> -a 3 -d chain


For questions, comments or suggestions feel free to contact Elodie Laine or Alessandra Carbone.


If you use JET2, please cite:

E.Laine, A.Carbone. The local geometry and evolutionary conservation of protein surfaces reveal the multiple recognition patches in protein-protein interactions, PLoS Computational Biology, 11(12), e1004580, 2015.

java -cp $JET2_PATH:$JET2_PATH/jet/extLibs/vecmath.jar jet.JET -c <default.conf> -i <struct.pdb> -o `pwd` -p AVJC -r local -d chain -n 10