Ligand refinement in nuclear hormone receptors

The above movie shows 100 accepted steps of Benzoxazin binding to mineralocorticoid receptor simulation performed by PELE using the default sample PDB and script parameters and generated using the traj_.1.pdb.gz file below. Please note that because of the random nature of the simulations performed by PELE, you may not get exactly the same result.

Files

• Control script: dyn.con (1.1 KB)
• PDB file: initial.pdb (322 KB)
• Native file: native.pdb (209 KB)
• Trajectory: traj_.1.pdb.gz (6.8 MB)

LIGAND BINDING REFINEMENT

What it does

The ligand binding refinement ready-made script has been designed to refine docking poses. Thus, the ligand is already in the binding site or in the immediate vicinity. It does an exhaustive search of possible binding modes within an RMSD range of to the initial structure.

How it does

The control file introduces a spawning RMSD distance with respect to the initial structure. Of course, only structures within this RMSD distance will be explored. Alternatives spawning options are detailed below. The algorithm then does small ligand translations combined with both small and large rotations. Every time the ligand escapes the distance criteria, the trajectory starts again from the initial structure.

At each iteration the algorithm performs: translation and rotation of the ligand, a move in a given normal modes, side chain prediction (including the side chains of the ligand) and overall minimization.

Changing parameters

We recommend playing around with different parameters that will adapt better the simulation to your specific system.

Spawning RMSD: a larger RMSD spawning distance will allow exploring a larger region of the space. However it will demand more computational resources. Again, only structures within this RMSD distance will be explored. Keep in mind that for large flexible ligands large RMSD changes might be necessary and that often the RMSD space does not correspond to the real displacement of the center of mass (COM).

Spawning distance: an interesting alternative is to spawn to a reference point in the space, determined by a set of Cartesian coordinates, a given atom, the ligand’s initial position, etc. This will require to pre-determine a position within the active site and that the ligand’s COM is initially within the spawning distance to this position! Every time the ligand escapes the distance criteria, the trajectory starts again from the stored structure where the COM is closer to the reference point. This option gives usually faster and better results. To use it replace the spawning RMSD line to:

	spawn point 1  XX YY ZZ  within 6.0 &
	

Where XX, YY and ZZ are the x, y and z coordinates of the reference point.

ANM type: besides projecting normal modes for the entire protein, we can select the relebelly option, where those modes with larger changes in the active site (within a user-defined distance) will be used. To use it:

	lanmanm relebelly yes &
	lanmanm anmrelerad 15.0 &
	

Where you select a distance (anmrelerad) where the ANM is being applied. The decision of which one to use depends if we believe that some global/local modes might affect the ligand binding in the active site.

Expected outcome

The outcome will include a series of structures with different OPLS binding energies. In all our test cases, the best (lowest) binding energies corresponded to the closest structure to the true native bound state. Typically we use 8-16 processors for 2-10 hours (depending on the ligand complexity, flexibility of the active site, etc.)

Importantly, when comparing poses from different ligands, however, we recommend using other scoring functions (such as Glide, etc). This is necessary due to the intensive nature of the OPLS binding energy. Typically, we cluster the results and score them by with Glide redocking.

Caution

In order to obtain good results check the parameterization of the ligand and that its initial position is within the spawning distance from the active site! Soon the server will include an interactive ligand parameterization tool.

References

For further reading, please check the performance of the refinement algorithms in:

1. Borrelli, K., Cossins, B.P. & Guallar, V. Exploring hierarchical refinement techniques for induced fit docking with protein and ligand flexibility [J. Comput. Chem 31:1224-1235 (2010)].

Application studies to real systems (in collaboration with experimental studies) can be found in:

1. Hosseini, A., Espona-Fiedler, M., Soto-Cerrato, V., Quesada, R., Pérez-Tomás, R. & Guallar, V. Molecular Interactions of Prodiginines with the BH3 Domain of Anti-Apoptotic Bcl-2 Family Members [PLoS ONE 8(2):e57562 (2013)].
2. Espona-Fiedler, M., Soto-Cerrato, V., Hosseini, A., Lizcano, J.M., Guallar, V., Quesada, R., Gao, T. & Pérez-Tomás, R. Identification of dual mTORC1 and mTORC2 inhibitors in melanoma cells: Prodigiosin vs. obatoclax [Biochem. Pharmacol. 83(4):489-96 (2012)].