Troubleshooting Common Issues

There are a number of areas where issues with FEP+ calculations can arise. Common problems and possible solutions are detailed here.

Poor convergence

Extend the edges by:

  • editing the jobname.edge file to include only the edges you wish to extend
  • running the extend_jobname.sh script

Poor similarity between perturbations

Add a ligand intermediate to the map.

Problematic results for elongated proteins

The default simulation box is constructed to minimize the volume. In this situation, a very elongated protein can have a box that is far from cubic, with the long axis of the protein aligned with the long axis of the cell. If the protein rotates, the ends could cross the boundary and interact strongly with the image of the protein in the neighboring cell. This can cause erratic results. The remedy in this case is to make the box cubic, which you can do by editing the input MSJ file.

Another situation can occur if the principal axis of the protein is aligned diagonally across the box. In this case the protein is not usually much elongated and the box is likely to be close to cubic. In this case, you can increase the buffer size to allow the protein to rotate. The increase would need to be about 70%, to allow sufficient space when the protein is aligned parallel to the cell axes.

Suspected ring flipping

In cases where ring flipping or a change in functional group orientation is suspected to occur when perturbing from one ligand to another, ensure that these ligands are not directly connected. In the absence of concrete information, a fictitious intermediate could be constructed and used in the map to connect the compounds in question. In the example below, a terminal methyl could replace the pyridine to connect CAT-17e and CAT-24. This would allow you to delete the edge between CAT-17e and CAT-24 as there is no assumption about the orientation of the pyrimidine.

Asymmetrically substituted rings

This can be achieved by adding in additional intermediates that do not contain the ring or by using Custom cores.

The torsion angle distribution plots that are provided for the analysis of an FEP+ edge enables you to estimate the extent of the sampling around the ring of interest. The observed rotamer states can guide decision-making about whether additional intermediates or custom cores should be used in follow-up simulations.

Custom cores

In cases like suspected ring flipping, asymmetrically substituted rings, or when you wish to swap entire fused-ring systems, you can readily increase the flexibility of the sampling by using custom cores. See Using a Custom Core in the FEP+ Panel for more information.

Subjobs failing

Subjobs can fail for a variety of reasons, including the rare set up errors or computing cluster issues. With a densely connected map, you can still get information out of the graph, so it is valuable to look at the results before resubmitting the job. If subjobs have failed, you can restart the FEP+ job with the restart_jobname.sh script. Ensure that you add in the SCHRODINGER variables and the OPLSDIR path to the restart script, for example

$SCHRODINGER/fep_plus -HOST hostname -SUBHOST subhost_name -JOBNAME jobname
-RESTART -checkpoint jobname-multisim_checkpoint -OPLSDIR oplsdir

Errors in the FEP+ Panel Status column

Hover over the yellow exclamation point to see the nature of the error in the tooltip. If there are missing torsions, run the Force Field Builder. This can take overnight, so it is important that your laptop does not go to sleep if you launch this job from the GUI. Once the calculation is complete, it is critical to save the parameters.

Ligand has slow degrees of freedom

Looking at additional Simulation Interaction Diagram reports or trajectories might help you to understand what went wrong and therefore what might address the issue. Situations that cause slower degrees of freedom include:

  • inadequate sampling during the 5 ns simulation
  • exchange of deeply buried water molecules
  • flipping of nearby protein residues
  • flipping of perturbed R-groups
  • large scale protein conformational change

A few strategies can be used to better sample slow degrees of freedom:

  • explicitly include crystallographic waters to allow buried waters to be adequately sampled
  • perform larger perturbations; this enables more room for the waters to move around in the presence of the smaller ligand
  • modify the REST region assignment to run with additional protein flexibility and thus improve protein and ligand sampling efficiency
  • add intermediates to force larger perturbations that would facilitate R-group flipping
  • run longer simulations by extending edges
  • dividing the series; if water molecules that are deeply buried in the binding pocket have slow degrees of freedom, consider dividing the ligands into subgroups that have the same number of water molecules in the pocket

Bad hysteresis

If you identify a problematic edge or ligand, this could indicate that an inconsistent binding mode was adopted for a ligand in one of its edges relative to another, or perhaps that the protein was unstable in one perturbation. Try removing the ligand or the edge and see if you can gradually reduce the number of cycles with large hysteresis. To quickly identify edges with bad hysteresis, in the Map tab, click Display perturbation properties and select Highlight bad perturbations. This will identify all edges that contribute to high hysteresis in red. You may need to re-import the map if you remove a ligand or an edge and later realize you want to keep it. After ligand/edge removal, check to see that the Predicted Error is reduced or fewer cycles have large hysteresis. In maps with bad hysteresis, once you have removed the edges or ligands that are problematic you can write out a new _out.fmp file for later.

High Bennett error

If you have Bennett Errors greater than 0.3 for a given edge, check if there is good sampling of ligand torsional angles. You can view these from the Simulation Interaction Diagram (SID) report. If torsions are not being explored, consider adding an intermediate between the two ligands being perturbed that would allow for free rotation. For instance, adding an intermediate that removed an asymmetric ring and replaced it with a methyl group.

Force field coverage

Force field coverage of torsional angle space is assessed within the FEP+ panel. Proper coverage of ligand torsions is critical to the success of an FEP+ calculation. Ligands should not be simulated in FEP+ if they do not have torsional angle parameters, otherwise there may be a significant loss in accuracy of the predictions.

Once the ligands are imported, a green check mark in the bottom right corner of the FEP+ setup panel, next to Total Ligands, indicates that the ligands have full torsional angle parameters.

If ligands need torsional angle parameters, you can generate them by selecting the Generate missing parameters with Force Field Builder option in the Generate missing parameters with Force Field Builder option

You can also change the custom force field used after the ligand health check is run in the setup panel, by clicking the ligand health icon.