Figure 2-1. Importing the input data for this tutorial.

1. Open Maestro and create a new project named FEP_BACE1.prj for this tutorial.
   • Don’t know how? See First steps in Maestro.
2. Download the tutorial zip file including input files and reference outputs here: https://www.schrodinger.com/sites/default/files/s3/release/current/Tutorials/zip/fep_bace1.zip
3. After downloading the zip file, unzip the contents in your Working Directorythe location that files are saved for ease of access throughout the tutorial.
4. Go to File > Import Structures.
5. Find and choose FEP_BACE1_tutorial.maegz from the tutorial files.

Figure 3-1. Tyr132 in the binding pocket.

We have provided two structures of the BACE1 protein to highlight the impact of receptor choice.

1. Ctrl-click (Cmd-click) to includethe entry is represented in the Workspace, the circle in the In column is blue 4DJV_4b_prepared and 4DJX_17h_prepared
2. Type L
   • Tyr132 has a different orientation between 4DJV and 4DJX
   • The binding pocket is more open with 4DJV
   • The phenyl group of the 4DJV ligand clashes with the 4DJX Tyr132 sidechain

Figure 3-2. Inspect ligand alignments.

Next, let’s look at the alignments for the set of ligands used for this tutorial.

3. Double-click the In circle to fix the 4DJV_4b_prepared receptor in the workspace.
4. Select the FepPoseBuilder_BACE1_subset group in the Entries.
5. Use the left and right arrow keys to step through the ligands and inspect their poses.

Note: You can either use the poses provided with the tutorial files for this tutorial or your own results from the Ligand Binding Pose Generation for FEP+ tutorial. Not all ligands from that data set are used in this tutorial to keep the computational costs of the FEP+ job reasonably low.

Figure 3-3. Duplicate 4DJV_4b_prepared into an Existing Group.

Finally, you’ll need to create a copy of the receptor structure minus the ligand.

6. In the Entry Lista simplified view of the Project Table that allows you to perform basic operations such as selection and inclusion, right-click 4DJV_4b_prepared
7. Choose Duplicate > In Place
   • A duplicated entry is added to this group
8. Double-click the new entry and rename it 4DJV_FEP
9. Includethe entry is represented in the Workspace, the circle in the In column is blue 4DJV_FEP

Figure 3-4. Select and delete the ligand from 4DJV_FEP.

Now, select and delete the ligand from the structure.

10. Type L
    • The Workspacethe 3D display area in the center of the main window, where molecular structures are displayed is zoomed to the ligand
11. Double-click a ligand atom
    • The whole ligand is selected
12. Right-click and choose Delete Atoms

Figure 4-1. FEP+ in Tasks.

1. Find FEP+ in the Tasks and open it.

Figure 4-2. Import structures from the Project Table.

2. Select(1) the atoms are chosen in the Workspace. These atoms are referred to as "the selection" or "the atom selection". Workspace operations are performed on the selected atoms. (2) The entry is chosen in the Entry List (and Project Table) and the row for the entry is highlighted. Project operations are performed on all selected entries both the 4DJV_FEP entry and the FepPoseBuilder_BACE1_subset group.
3. For Import structures or perturbation map from, choose Project Table (14 selected entries)
4. Click Import
   • Health checks for the inputs are performed (see below)
5. When the import is finished, click Relative

Figure 4-3. The protein reliability results for this receptor.

Optional: For the receptor, clicking on the warning will open the Protein Reliability Report showing various quality metrics.

6. Click the warning sign for the receptor
7. Click the Peptide Planarity heading to see the issues.
8. Click the rows to see where the offending residues are located.

The only issues with this receptor are a few minor deviations from peptide planarity outside the binding site, which you can safely ignore.

Figure 4-4. A glance at the Force Field builder

Optional: If you’re working with your own set of ligands, some torsions may be incompletely parametrized in the OPLS force field, causing the yellow warning sign.

In that case, you should run Force Field Builder on the ligands (either manually or as part of the FEP+ job) before proceeding with the calculation.

None of the ligands used in this tutorial should be missing torsion parameters.

Figure 4-5. Add Affinity Data.

9. In the FEP+ panel, click Affinity and choose Experimental Data
   • The Choose Affinity Property panel opens

Figure 4-6. Choose dG. exp (user) as the Affinity Property.

10. Choose dG.exp (user)
11. Next to Affinity units, choose ΔG (kcal/mol)
12. Click OK

Note: When working on your own ligand set, you can either add the reference values as properties to the ligand entries, import them from a CSV file using this panel, or add them manually in the FEP+ panel by double-clicking the cell in the Exp. Affinity column for each ligand.

Figure 4-8. Generate Map.

We will choose bias ligands based on the protein structure used and known affinity: CAT-4b is the cognate ligand for the receptor structure we are using, and CAT-17h has a binding affinity in the middle of the range of known binders. Biasing to the best known binder may introduce a selection bias, distorting the statistics.

13. Click the Map tab
14. In the Bias Column, choose CAT-4b and CAT-17h
15. Click Generate Map
    • The Map Options panel opens

Figure 4-9. Map topology options.

We will start from an optimal map and add further edges to allow more leeway in identifying and excluding problematic ligands and perturbations.

16. Next to Map topology type, choose Optimal
17. Click Generate Map
    • This job will take less than one minute  

Note: Zoom and translate the map with the same mouse actions as the Workspacethe 3D display area in the center of the main window, where molecular structures are displayed. Click and drag to move the ligands in the map.

Figure 4-10. Show similarity scores and view the atom properties of the ligands.

You should make sure that edges are only drawn between sufficiently similar ligands.

18. Click Display perturbation properties
19. Choose Similarity scores
    • Scores are shown along the map edges
    • Darker edges indicate higher similarity

Note: You can also view the FEP+ protocol (default, charge perturbation, core-hopping, or fragment linking) for each edge. The appropriate protocol is chosen automatically when the edge is drawn. See here for more information.

Figure 4-11. Add edges to the map.

You can now manually add more edges to the map.

20. Click Add new connection (on the right-hand side below the map)
21. Click CAT-13f and CAT-13o
    • An edge is added
    • The similarity score is shown
22. Repeat steps 29 and 30 to ensure edges exist between the following ligands:
    • CAT-4i and CAT-4d,
    • CAT-13i and CAT-17e,
    • CAT-13d and CAT-13b,
    • CAT-17f and CAT-17e,
    • CAT-4c and CAT-4d.
23. Optional: Click the Auto-layout button (next to Add new connection) to make the map more legible.

Note: The “_1” and “_2” suffixes for some ligands denote which of the available Epik states of that ligand was kept by the filtering stage of the FEP+ Pose Builder. As we are only using one state per ligand, we will ignore these suffixes for the rest of this tutorial when referring to specific ligands.

Figure 4-13. View the hot atoms of an edge in Atom Properties.

You can select an individual edge to see additional information about the perturbation.

24. Click on an edge
    • Atom Properties of the two compounds are shown
25. Choose Hot atoms (complex leg)
    • The REST regions of the ligands are highlighted.

Note: Click an empty area in the map to clear edge selection

Figure 4-14. View the FEP+ map info.

26. Optional: Click on Info in the bottom right to get a quick look at the properties of information in the map (receptor health, ligand health, similarity score, etc.). We will ignore the warning about missing the FMPdb since this map has not run.

Figure 4-15. Rearrange the map layout and change the Job name.

You can now name your job and specify which computational resources to use.

27. Change the Job name to fep_BACE1_4DJV
28. Click Run Settings (cog)

Figure 4-16. Set hosts (top) and write the FEP+ job (bottom).

29. Choose your CPU Host and GPU Host

Ensure Maximum simultaneous subjobs is set to 0. This removes the limit on the number of subjobs, so they are all submitted to the subjob host queue. If you do not have license checking enabled, set the number of subjobs to ensure that you do not exhaust your licenses.

You can now either run the job or write out the job files to manually transfer them to a remote HPC.

30. Optional: Click Run.
    • This job requires several hours and  significant GPU resources to run.
    • Instead of running the job yourself, we recommend continuing with the results in the tutorial files.

Figure 5-1. FEP+ in Tasks.

1. Find and open FEP+ in Tasks.

Figure 5-2. Import perturbation map.

2. For Import structures or perturbation map from, choose File
3. Click Browse
4. Choose fep_BACE1_4DJV_out.fmp and click Open
5. When the import is finished, click Next
   • The tables and map are populated

Figure 5-3. Rearrange the data.

For a first overview of the results, we recommend checking the range of predicted binding affinity errors.

6. Right-click the Pred. Binding(∆G) column header and choose Sort by Error
7. Left-click Pred. Binding (∆G)
   • The table is reordered with the highest error values (in kcal/mol) at the top.  

The error values for the predicted ∆G are   values are the error bars of the FEP+ calculation, not the error of the prediction w.r.t. the experimental reference. Predicted errors of <0.5 kcal/mol are good, values >1 kcal/mol indicate that there are some inconsistencies in the map. In this case, the ligand-wise predicted ∆G errors all are well below 0.5 kcal/mol.

8. Click Plot
   • The Correlation Plot (FEP+) panel opens

Figure 5-4. Correlation plot of FEP+ results.

Any outliers in this plot are ligands you should keep in mind while performing further analysis steps.

9. Hover over the worst outlier data point
   • A tooltip with ligand information appears
10. Optional: Click Save As to save a .png of the Correlation Plot to your Working Directorythe location that files are saved.

Figure 5-5. Analyze edges.

In our case, the edgewise RMSE is at 1.16, but the pairwise RMSE is at 1.55, so we’ll want to investigate further. You can now go back to the map to check on the performance of the individual edges for potential convergence issues.

11. Close the correlation plot.
12. Switch to the Map tab in the FEP+ panel.
13. Click Display perturbation properties
14. Choose Predicted ddG (corrected) and Experimental ddG.

Note: You can zoom in and out of the map with your scroll wheel. Depending on the zoom level, ligands will either appear as structures or numbers.

Figure 5-6. Examine the hysteresis.

15. Click the Hysteresis tab
16. Click the Hysteresis column header twice
    • The table is reordered with the bad hysteresis cycles at the top
17. Click through the cycles
    • The selected cycle is highlighted in the map.

Figure 5-7. Optional: Display bad perturbations.

First, we can show perturbations which are involved in cycles with bad hysteresis in the map.

18. Click Display perturbation properties and choose Bad perturbations.
    • Edges with poor convergence are colored and labeled.

Note: See Identifying Unconverged Edges in FEP+ Jobs for details on when an edge is labeled a bad perturbation.

Figure 5-8. Analyze the edge.

You can analyze individual edges to see what might be causing poor convergence.

19. Right-click the bad perturbation edge connecting ligands 12 (CAT-13o_2) and 9 (CAT-13f_1).
20. Choose Analyze.
    • The Analysis: CAT-13o_2 – CAT-13f_1 panel opens.

Figure 5-9. Edge analysis summary

The initial summary shows you the two ligands and allows you to highlight e.g. the perturbed atoms, the common core, the REST region, the atom-wise RMSF, and the atom mapping between the two ligands. This can help you understand which part of the ligand might be causing issues.

Note: You can generate a PDF report of the analyses available in this panel from the button in the lower right corner. The plots in the report contain additional notes explaining what is shown and providing guidelines for interpretation. You can generate the report yourself or see the FEP_SID_report_cat13o_2-cat13f_1.pdf file in the tutorial files for an example.

Figure 5-10. Analysis of the torsional sampling.

You can now check whether the torsions in the ligands were well-sampled during the simulation, as this can hamper convergence.

21. In the Analysis panel, click the Ligand Details tab.
22. Click the Rotatable Bonds tab.
23. Hover over the torsion sampling graphs to highlight the atoms involved in the corresponding torsion.
    • The sampling of the green and purple torsions for the CAT-13f ligand seems to be limited in both the solvent (hollow bars) and complex (filled bars) legs.

Note: The gray bars indicate the initial angle of the rotatable bond in the input structure. The strain in kcal/mol is reported below the plot for the complex and solvent legs. See more details here.

Figure 5-11. Analysis of the convergence.

24. Click the Convergence tab
    • This perturbation may benefit from a longer simulation time in both the solvent and complex legs, as both the Reverse and Sliding Time do not appear stable.

Note: The y-axis of the convergence plots will adjust to fit the scale. Pay attention to the y-axis values, as good convergence may appear poor if the scale is very small and vice versa.

Figure 5-12. Analysis of the REST Sampling.

You can check the sampling for both the solvent and complex legs.

25. Click the REST Sampling tab
26. Click Complex.
    • While the sampling is mostly uniform for the solvent leg, in the complex it is not uniform.
27. Close the Analysis panel

Figure 5-13. Delete the edge.

Let’s delete both bad perturbations from the map.

28. In the FEP+ panel, once again right-click the edge connecting ligands 9 (CAT-13f_1) and 12 (CAT-13o_2).
29. Choose Delete.
30. Also delete the edge connecting ligands 10 (CAT-13d) and 11 (CAT-13b_1).
31. Optional: Check the Hysteresis table again.
    • The two perturbations you removed were responsible for all of the bad hysteresis cycles.

Note: You can undo the removal of an edge or a node using the controls below the map.

Figure 5-14. The Correlation Plot is improved.

Deleting bad perturbations often reduces the error bars on the predicted ∆Gs.

32. Click Plot
    • The error bars of the data points are reduced.
33. Close the Correlation Plot

Figure 5-15. Reorder by predicted FEP.

The Analysis tab of the FEP+ panel provides an overview of the relevant quality metrics for the individual edges.

34. Click the Analysis tab
35. Click the FEP column twice
    • The table is reordered to show the predicted Bennett Free energy difference (ΔΔG) in descending order.
    • Errors for predictions are also shown.  

Note: If the error for an edge in the FEP column is >0.3 kcal/mol, this is a sign that there is still an inconsistency in the map. Please see Troubleshooting Common Issues for more information.

Figure 5-16. Hover over descriptions to see more information.

The energy convergence, ligand RMSD, REST exchange, and cycle closure correction convergence columns provide another possibility to identify issues for individual edges.

36. Hover over the Good, Fair, and Bad descriptions to get more information about the individual metrics.

Note: The image on the left shows a representative example from a different FEP+ job.

Figure 6-1. Activity Cliffs tab of the FEP+ panel showing positive correlation.

First, we’ll see how this looks for a ligand where there’s a positive correlation between similarity and potency.

1. In the FEP+ panel, switch to the Activity Cliffs tab.
2. Set the reference ligand to CAT-17f
   • The data points cluster towards the lower right – the more similar a ligand is to this reference, the more potent it is.
3. Click on the data point for CAT-24 (see figure)
   • The chosen ligand is highlighted in the table.
   • The small perturbation from the nitrile to the alkyne leads to a large increase in affinity.

Figure 6-2. Activity Cliffs tab of the FEP+ panel showing negative correlation.

You can compare this to a ligand where the correlation is inverted (and less-pronounced).

4. Set the reference ligand to CAT-4d.
   • Now, the ligands that are most similar to the reference do not show favorable potency changes.
5. Click on the data point for CAT-4c (see figure)
   • Moving the substituent on the last phenyl ring to the para position makes the binding affinity worse by ~2 kcal/mol.

Figure 6-3. Display the Complex Trajectory.

Let’s check the complex leg trajectory of the edge connecting CAT-4c and CAT-4d.

37. In the Complex Trajectory row for the CAT-4c <--> CAT-d edge, click 5.0 ns
    • A new group with entries for each ligand are added to the Entry Lista simplified view of the Project Table that allows you to perform basic operations such as selection and inclusion

Note: If there were no edge between this pair of ligands in the map, you could also use the trajectories from other edges containing either CAT-4c or CAT-4d.

Figure 6-4. Load and play the trajectory.

38. Click the T associated with the CAT-4d entry and choose Load Trajectory.
39. Click Play in the Trajectory Player to play the trajectory.
40. Repeat the previous two steps with the CAT-4c trajectory.

See the Introduction to MD Trajectory Analysis with Desmond tutorial for an overview of the analysis tools available in the trajectory player.