Get   for     ? 
 Site search     ? 
CAPRI: Critical Assessment of PRediction of Interactions
 
  
   Call For Targets
   Meeting Report
   Evaluation Meeting
Registration
   Evaluation Meeting
Program
   ROUND 1
   ROUND 2
   ROUND 3
 
 MSD  CAPRI: Critical Assessment of PRediction of Interactions

Third community wide experiment on the comparative evaluation of protein-protein docking for structure prediction

Hosted By EMBL/EBI-MSD Group

CAPRI Target 09 evaluation results

Raúl Méndez, Raphaël Leplae and Shoshana J. Wodak.
SCMBB Université Libre de Bruxelles, Cp 263, Brussels, Belgium.
Monday May 5, 2003.
e-mail: raul@ucmb.ulb.ac.be
shosh@ucmb.ulb.ac.be

The evaluation results of the CAPRI target 09 predictions are stored in different directories depending on the criteria that have been used. In the following the directories and their contents are briefly described.


Information.

Directory Information contains the information about target 09, that was used in the evaluation and scoring. It contains the following files (file names are given in bold): Please consider all information concerning the targets as confidential until the X-ray structures are released by their authors.

  • capri_09_xray.pdb: the crystal structure of the target (target 09) in PDB format: LicT homodimer (subunits A and B).
  • capri_09_xray.A.contres: list of residue contacts in the target between both subunits (A and B) of the LicT. For this target we mantain the notation "Receptor" and "Ligand", althought it's arbitrarily assigned.
  • capri_09_xray.A.intres: LicT interface (among the two subunits) residues.
  • capri_09_xray.A.d: list of clashes in the target interface.
    Final Summary

    File target 09 Final Summary. summarizes all the information about the target 09 evaluation in the same way as the corresponding summary file for target 08. It looks like that:

    
    
    PREDS       fnat      fnon-nat       fIR           INTERFACE RES.(OP) IA(A2) THETAANGLE   DIST  Nclash L_rmsd   I_rmsd
                                 Ligand  Receptor  Ligand   Receptor  
    
    T09_P03.10.B 0.200 0.919 0.729 0.688 0.632 0.649 8219.0 39.7 11.231 20 12.671 9.601 T09_P03.3.B 0.200 0.919 0.688 0.708 0.637 0.642 8339.3 39.8 10.428 4 11.979 9.569 T09_P03.2.A 0.100 0.948 0.625 0.604 0.670 0.667 6894.2 39.1 13.368 7 14.544 9.113 T09_P03.6.B 0.100 0.959 0.604 0.625 0.691 0.681 8146.4 31.2 12.458 13 13.330 9.675 T09_P03.4.B 0.078 0.971 0.625 0.583 0.670 0.708 8362.6 40.4 15.648 36 16.868 9.375
    .
    .

    Again T09_P03.10.B means participant 03, prediction 10 for the target 09, Ligand interface B (although, here the label for the interface it's totally arbitrary). For simplicity we call after aligned all sequences and sorted by lenght, we divided into two sets, and the sequences belonging to the first one (the one with the longest sequence) we call them "Receptor" and the rest "Ligands". But from now on we will call them Ligand and Receptor, respectively.

    Column 2 gives the fraction of predicted contacts over native. This fraction is computed as the number of contacts in the prediction that match the contacts in the target, divided by the number of contacts in the target. As for target 08, 2 residues are considered as being in contact if at least one atom of one residue is within 5Å of an atoms of the other.

    Colum 3 gives the fraction of non native predicted contacts (over prediction). This fraction is computed as the number of contacts in the prediction that doesn't match the contacts in the target, divided by the number of contacts in the prediction. This number accounts for the real efficiency of the prediction in term of contact: as bigger is the predicted interface as higher the probability of predict native contacts.

    Columns 4 and 5 list the interface residues ratios over native (fIR). Column 4 give the ratio between the residues of the Ligand that are part of the interface in the prediction, over the Ligand residues that are part of the interface in the target. The 5th column gives the same information for the Receptor moiety. All the interface residues lists are generated using the BRUGEL package.

    Columns 6 and 7 lists the interface residue ratios over prediction. They are analogous to columns 4 and 5 but now dividing the number of residues in the prediction found in the target over the total number of provided residues at the predicted interface.

    Column 8 lists the interface Area (in Å2), calculated as IA = ASA(ligand) + ASA(Receptor) - ASA(complex). For ASA (Accessible Surface Area) calculations we used Conolly the (Conolly, M.L. 1985, J. Mol. Biology., 107, 1118 - 1124) and Richmon (Richmon, T. J. 1984, J. Mol. Biol., 178, 63-89) algorithm, implemented also in the BRUGEL package.

    Column 9 lists the rotation angle (Theta angle) necessary to fit the Ligand molecule in the predicted complex to that in the target, as per capri_09_xray.pdb. To compute this angle, we first perform a rigid-body fit (Kabsch, 1978, Acta. Cryst. A. 34, 827-828) of the Receptor subunit in the predicted complex, to the Receptor subunit in the target.

    For that particular target, the conformational change between the provided unbound coordinates (coming from the same heterodimer, but the double mutant H207D and H269D as it is in Targets_description) and bound is huge. So only the fragment from the domain which changes the least are considered (in both prediction and target). See below for details.

    After this first fit, a second fit is performed so as to superimpose the predicted Ligand molecules onto its closest counterpart in the target structure (capri_09_xray.pdb closest). The rotation angle corresponding to this second fitting is the listed theta angle.

    Column 10 lists the distance (in Angstroms) between geometric centers of predicted and target Ligand molecules before the second fit. The distance between the geometric centers together with the Theta angle gives an idea of the global position of the Ligand in the prediction relative to the position in the target.

    Column 11 lists the number of clashes Nclash between the Ligand and the Receptors for each predicted complex. Clashes are computed between heavy atoms within 3 Å . In the detailed information you can find the close contact pairs classified into three categories: from 0 to 1, from 1 to 2 and from 2 to 3 Å.

    Columns 12 and 13 list the RMSD's (Root Mean Square Deviation) values in Å . Column 11 list the RMSD values calculated between the Ligand's backbones once the Receptors are superimposed (Ligand RMSD or L_rmsd). Column 12 contains the rsmd's when sumperimposing the backbones of the residues at the interface (Ligand + Receptor) on the prediction upon the counterpart in the target. Residues at the interface (Interface RMSD or I_rmsd) are re-defined here, as residues in the target having at least one atom within 10 Å of an atom of the other molecule. The equivalents for those residues in the predictions are considered as to be in the interface to sumperimpose. For all the RMSD calculations we consider the same molecular fragments as for the fits, but in the case of the interface RMSD's, restricted to the residues at the interface, according to this new definition.


    Contact List

    Directory ContactList contains one file per predicted interface, with information on the residue-residue contacts in the predicted versus the target complexes

    As an example the file T09_P03.10.B.highlighted is illustrated in part:

    HIGHLIGTHED CONTACT LIST FOR T09_P03.10.B
    Number of Contacts = 223 Matching List1 = 18/90



    B3    MET - A18   MET 1
    B4    GLU - A42   VAL
    B5    LYS - A22   GLU 1
    B5    LYS - A38   ASP
    
    .
    .

    Each predicted contact that matches the target contact list is highlighted with a number indicating the reference list is matching. For this round "1" refers to the only possible list.


    INTERFACE_RESIDUES_HIGHLIGHTED

    Directory InterfaceResidues contains one file per predicted interface, with information on the residues forming the Receptor - Ligand interface in the prediction and how well they match those in the target interfaces.

    The information contained in each file is illustrated by an example, T09_P03.10.B.highlighted

    HIGHLIGHTED INTERFACE RESIDUE LIST FOR T09_P03.10.B
    N_res_Ligand = 95 N_res_Receptor = 94 Match Ligand in List1 = 35/48 Matching Receptor in List1 = 33/48


    Ligand LIST
    
    B2    ALA    0.300 1
    B3    MET   34.387 1
    B4    GLU    0.001
    B5    LYS   89.278 1
    B6    PHE  164.606 1
    
    .
    . Target LIST A3 MET 28.070 1 A5 LYS 77.874 1 A6 PHE 153.545 1 A7 LYS 75.684 1 A8 THR 43.529
    .
    .

    Each time a residue of the Ligand or Receptor in the predicted interface matches one of the interface residues in the target list, it is highlighted with the number of the corresponding target reference list, 1 again stands for the number of the unique list.

    Note that interface residues list files and contact list ones are named the same (i.e. T09_P03.10.B.highlighted) but they are in different directories and their contents are completely different.


    FITTING_SUMMARY

    Directory FittingSummary contains one file per predicted interface, with information on the results of fitting the predicted complex over the target complex. The information contained in each file is illustrated by an example, file T09_P03.10.B.fitting.summary

    Fitting of A prediction receptor Subunit onto X CAPRI receptor Subunit
    Rotation Matrix:
      -0.41916   0.10496   0.90182
      -0.68750  -0.68546  -0.23976
       0.59300  -0.72050   0.35948
      Translation vector      9.413     5.806    84.043
    Fitting Ligands, B onto Y
    Theta angle = 39.67
    Distance between geometric centres = 11.23113
    

    As for the evaluation of target 08, we give the information about the first fit (rotation matrix and translation vector including which subunits are involved), the distance between predicted ligand and Capri ligand after this first fit (considering just the fragment that is fitted in the second fit) and the Theta angle of the second fit.

    This target shows big conformational changes (unbound - bound). Unbound molecules come from the double mutant H207D and H269D (see above for the reference). Replacement of the two mutant Aspartic residues by wild type Histidines seems to have a dramatic effect mostly on PRD1 domains. For this reason we used the program Sofist, to obtain the most structurally conserved segments among the unboud and bound molecules. References about this program can be found at the same site, and also at Sofist_ref.

    Using this program and after the corresponding multiple sequence alignment, the longest fragment least changed for the first fitting is formed by residues 71-84, 89-94, 96-112 and 116-164 on the targets while the second fit was made considering the same residues in the ligands, as they are homo dimers molecules. As we already pointed out, here "Target" and "Ligand" terms are completely arbitrary but don't affect the final results, since the relative shiftment are equivalent no matter which subunit is taken as a reference. In order to be consistent,the distance between geometric centres was calculated taking into account only this ligand fragment.

    Note that in order to not confuse chain ID's between target and predicted coordinate sets, the chain ID's in the target (capri_09_xray.pdb) were renamed as follows:

    B to X
    for the Receptor subunit
    A to Y
    for the Ligand ligand subunit.


    FITTED PDB

    Directory FittedPDB contains the files with the coordinates of the predicted and target complexes superimposed, following the first fit, in which the Receptor subunits have been superimposed (using the listed rotation matrix and translation vector).


    CLOSE_CONTACTS

    Directory CloseContacts contains one file per predicted interface with information on the clashes in each predicted interface.

    For example part of file cc.T09_P03.10.B.d looks like that:

    Ligand Atom         Receptor Atom           Distance
    
    --
    B 6    .PHE.CE2     A 41   .TYR.CD2         2.42
    B 6    .PHE.O       A 42   .VAL.CA          2.46
    B 8    .THR.CA      A 7    .LYS.O           2.79
    B 222  .LYS.NZ      A 171  .GLU.OE1         2.81
    B 5    .LYS.O       A 42   .VAL.CG2         2.82
    B 6    .PHE.C       A 42   .VAL.CG2         2.82
    
    .
    .

    As in the evaluation of target 08, the list of clashes is segregated into clashes between 0-1, 1-2 (no contacts in this case and 2-3Å. Empty files means, no close contacts found.