In this example, a 6ns MD trajectory is used for identification of transient binding pockets in aldose reductase (AR)


1. Overview :

There are three flexible pocket regions in human AR denoted as a, b, and c. Sub-pocket a is occupied by tolrestat, TOL, and the inhibitor compound, IDD594, in crystal structures 1ah3 and 1us0, respectively; the ligand citric acid (CIT) occupies sub-pocket b in structure 2acu of the Y48H mutant of AR; the two molecules of the drug alrestatin, ALR, (both present in the crystal structure 1az1 of the C298A/W219Y mutant of AR) stack with each other with one occupying sub-pocket a and one occupying sub-pocket c (see Fig. 1). In this example , we will analyse a 6ns MD simulation of the Y48H mutant of human aldose reductase (AR)




graphics1
Fig.1: PDB structures of wildtype and mutant AR: 1us0 , 1ah3, 2acu, and 2az1 are shown in blue, red, black, and green; The ligands have residue names (A) TOL (1ah3), (B) CIT (2acu), and (C) ALR (1az1). For more details see here.

2. To run this example, you need the following files and parameters:

  • Ref_in.pdb the reference structure (2acu pdb file, containing only the protein with H atoms added)
  • Md.pdb a trajectory in PDB format that consists of 60 snapshots from an MD simulation (written at 100ps intervals)
  • Binding pocket center defined by x, y, z coordinates in Angstroms: 66.9, 37.56, 89.35
  • Radius defined in Angstrom: 8

3. Simulation Procedure and Results

  • Step 1: Uploading the reference structure and defining the input parameters
  • You have to upload the reference structure and to specify coordinates of the pocket center and the effective pocket radius (for this, choose Manually defined center).

    You also have to click Upload PDB or DCD trajectories at the next step before you go to the Next Step (See Fig.2).




Fig.2: Your webpage will look like this.


  • Step 2: Uploading trajectories
  • On the next page Pocket analysis will be started with the following parameters, you must be able to see a JSmol visualization of the reference structure and an identified binding pocket. You can then upload the trajectory (or trajectories) and launch TRAPP-structure/analysis. Since no method for generation of new structures was selected in the previous step, only analysis of the uploaded trajectory(or trajectories) will be done. (See Fig.3)



Fig.3: Your webpage will look like this.



  • Step3 : Checking input data and starting simulations
  • On the page titled Pocket analysis will be started with the following parameters: - Press the button Launch TRAPP structure/analysis.

  • Step4: Viewing simulation results of TRAPP-analysis
  • When the job has completed you can open a page with the preliminary results of the TRAPP-analysis procedure. Note, that at this step the default parameters of the TRAPP-analysis are used (backbone atoms are used for RMSD calculations and a fast hierarchical clustering of the binding site conformations is carried out with a threshold of 3 Angstroms).

    Use the link View Analysis to see a summary of the simulation results, which includes:

      (i) A plot showing the RMS deviation of each residue from the reference structure in each snapshot of the trajectory (you can also download the data by clicking on the sign placed under the image). The regions of residues 218-219 and 122 are the most flexible with a deviation of up to 7 Angstroms from the reference structure. (see Fig.4)

      (ii) Clustering results: a distribution of snapshots over clusters. (See Fig.5 and Fig.6)


    Using the link List of Cluster Representatives you can download PDB structures of cluster representatives.


Fig.4: RMS deviation



Fig.5: Clustering results log file



Fig.6: Clustering results

  • Step 5: Re-run TRAPP-analysis using k-means clustering
  • You can change the parameters and re-run TRAPP-analysis. In particular, you can run a more accurate clustering; use geometric centers of the binding site residues instead of only backbone atoms to compute RMSD; change RMSD in clusters; or include additional residues in the binding site list. For this, choose the button Re-run TRAPP-Analysis and change the parameters on the next page.

    In this example, we add the residues 120, 121, 123, 124, 216, 217, 220, 221, and 222 to the binding site and choose k-means clustering.


    The results will be found on the "View Analysis" link. Binding site motions were found around the residues 48 and 217-222 (corresponding to flexible regions B and C in the crystal structures) and around 120-122. The analysis was done using backbone atoms to compute the RMSD (See Fig.7).
    When geometric centers of residues are used for clustering (See Fig.8) some fluctuation of residues 300 and 301 (sub-pocket a) appears.


Fig.7: RMSD plot (link "View Analysis")



Fig.8: RMSD plot (link "View Analysis")


Fig.9: Results of clustering (link View cluster representatives in JSmol): 4 clusters found. This shows the distribution of snapshots over the clusters.



Fig.10: Results of clustering (link View cluster representatives in JSmol): 4 clusters found. Summary of clustering.



Fig.11: Results of clustering (link View cluster representatives in JSmol): 4 clusters found. RMSD of cluster representatives from the reference structure.



Fig.12: Results of clustering (link View cluster representatives in JSmol): 4 clusters found. JSmol visualization of representative structures from the four clusters along with the reference structure. The red cartoon is the reference structure, the binding site residues are shown by ball/sticks and the representative structures of the 4 clusters can be viewed as models 2.1-5.1.

  • Step 6: Running TRAPP-pocket and analyzing the simulation results
  • In this example we click run PCA and run Clustering. Analysis will be done at the pocket occurrence of (%) of 10, 30, 65 then click Launch TRAPP-pocket.

    The results of TRAPP-pocket simulations can be viewed using the link: USER-TRJ Results: Trj-md to an HTML file that uses JSmol for visualization of transient pockets. (View simulation result in JSmol ).

    All three transient regions (A,B,C) are observed in at least 70 % of snapshots:
    • sub-pocket a is opened due to motion of C298 inside the binding pocket (observed in the reference structure);
    • sub-pocket b consists of opening (red) and closed (blue) parts relative to the reference structure that corresponds to changes in the position of H48 opening one pocket part and closing another;
    • sub-pocket c appears due to a loop motion (W219).

    Opening and closing of transient sub-pockets along the MD trajectory can be analyzed using the Pocket Characteristics tab.

    Here all transient sub-pockets are split into separate pockets (that are not connect with each other) and the large pockets are once more split into compact sub-pockets, then they overlap between pockets in each snapshot and the transient sub-pocket is computed and represented as a matrix.

    For example at a pocket occurrence of 65% (see plot ), one can see that sub-pocket c is mostly open because the loop moves opening a pocket (pocket 2, second column in the overlap matrix). (See Fig.13)

    Sub-pocket a is opened and closed periodically (due to the C298 side-chain rotation, see the third column (pocket 3) in the overlap matrix). This is shown at 30% occurrence of transient pockets that disappear. (See Fig.14)

    The main changes in the region of sub-pocket b appear in the last half of the MD trajectory (sub-pocket 2). (See Fig.14)


Fig.13: Overlap matrix at a pocket occurrence level of 65% (see plot). Sub-pocket c (pocket 2, second column) is mostly open because the loop moves opening a pocket.



Fig.14: Overlap matrix at a pocket occurence level of 30% shows that sub-pocket a (pocket 3) is opened and closed periodically (due to the C298 side-chain rotation, see the second column of the overlap matrix). This is shown at 30% of the tranisent pockets that disappear.



Fig.15: HTML page : Transient regions that appear in at least 70 % of snapshots



Fig.16: HTML page : Transient sub-pockets at the pocket occurrence level of 30%.