Ligand migration in nuclear hormone receptors
The above movie shows 367 accepted steps of Benzoxazin escaping out of the binding site of mineralocorticoid receptor performed by PELE using the provided PDB and control script and generated using the traj_.4.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.
What it does
The ligand migration ready-made script has been designed to force the ligand motion to previously decided regions of the space. It can produce, for example, entrance and exit pathways from binding sites.
How it does
As in the ligand refinement script, the control file introduces a spawning criterion. Thus, it is designed to use several processors that will share information towards a common goal. By default we use spawning to a reference fix point in the space (X,Y,Z) with a 4 Å distance. This means that if the center of mass (COM) of the ligand, for processor A, is 4 Å further away from the best (closest) structure produced previously (from any processor), then processor A will abandon its coordinates and receive the best (closest) ones. In the control file this is done by:
spawn point 1 X Y Z lt 4.0 &
Where X, Y and Z should be the real coordinate values that we are spawning to, and 4.0 is the spawning distance. Changing lt (lower than) for gt (greater than) will spawn away from a given point, allowing us to map exit pathways. We can also move away/towards a defined atom:
spawn atom 1 A:100:_CA_ gt 3.0 &
Where in this case we ask to move away from the alpha carbon of residue 100 in the A chain, with a 4 Å spawning. We can also force the migration of the ligand by asking an increase/decrease in the ligand’s RMSD:
spawn ini_rmsd_lig 1 heavy 0 gt 3.0 &
Where we aim for an increase in the RMSD of all heavy atoms (including 0 Å around it) from the initial ligand structure. Similarly:
spawn rmsd_lig 1 heavy 5 lt 3.0 &
Brings the ligand close to a native structure (thus you need to load a native structure with a ligand bound to it!). Notice that here we have also included heavy atoms 5 Å around the ligand.
At each iteration the algorithm performs: translation and rotation of the ligand, a move in a given normal mode, side chain prediction (including the side chains of the ligand) and overall minimization.
We recommend playing around with different parameters that will adapt better the simulation to your specific system.
Spawning distance: here the main parameters involved are the spawning type and its value. A small spawning 1~2 Å, will produce a quick migration but which might be too biased. We typically perform several simulations increasing the spawning value and check for differences. Obviously, for very large spawning distances the processor become independent.
Multiple task: If you want to compare pathways, then you might need to converge the energies (total or binding energy –the binding energy tends to converge more quickly) along the pathways. For this you will need to perform several simulations where you go back and forth a given pathway (usually more than two complete passages are required). A good way to perform this is by adding multiple tasks to the control file, for example:
task & spawn point 1 1.2 -3.5 20.4 gt 4.0 & exit point 1 1.2 -3.5 20.4 gt 15.0 & task & spawn point 1 1.2 -3.5 20.4 lt 4.0 & exit point 1 1.2 -3.5 20.4 lt 2.0 & task & spawn point 1 1.2 -3.5 20.4 gt 4.0 & exit point 1 1.2 -3.5 20.4 gt 15.0 & end_task &
Where we first move away from the reference point (1.2,-3.5,20.4), exiting when we reach a distance of 15 Å away. Then we go back to the point and a then a third and last iteration moving again away. In this way we can also build complex exit pathways with variable geometries…
Please, also read carefully the parameters in the free ligand exploration case since they also might apply here.
The outcome will describe structures and energies along predefined migration pathways. Here the trajectories should be considered as the ligand evolution. You should keep in mind, however, that the spawning will introduce large jumps in a given processor. Thus the trajectories have to be analyzed collectively. The quickness of the simulation depends on the spawning distance.
Typically we use 8-16 processors for ~10-24 hours to complete a search (depending on the size and ligand complexity length to explore, etc.).
Play around with the spawning distance. Very small spawning distances might create unrealistic paths and could also bring the ligand to no-exit “alleys”. We also advice to use distance spawning rather than RMSD spawning since big jumps in RMSD space might not involve migration along an exit/entrance path.
Several tests and application studies have been performed:
- Migration of carbon monoxide into hemoglobin:
- Lucas, F.M. & Guallar, V. An atomistic view on human hemoglobin carbon monoxide migration processes [Biophys. J. 102:887-96 (2012)].
- Entrance of alcohol ligands in apo-oxidase
- Hernández-Ortega, A., Borrelli, K., Ferreira, P., Medina, M., Martínez, A.T. & Guallar, V. Substrate diffusion and oxidation in GMC oxidoreductases: an experimental and computational study on fungal aryl-alcohol oxidase [Biochem. J. 436:341-50 (2011)].
- Ligand migration in truncated hemoglobin
- Guallar, V., Lu, C., Borrelli, K., Egawa, T. & Yeh, S.-R. Ligand migration in the truncated hemoglobin-II from Mycobacterium tuberculosis: the role of G8 tryptophan [J. Biol. Chem. 284:3106-16 (2009)].