Identifying Unconverged Edges in FEP+ Jobs

Sometimes there are a few edges that have not converged or have not been correctly set up. These edges will contribute to high hysteresis. The following procedure can be followed to identify these bad edges.

1. Open the FEP+ Panel and import the perturbation map.

2. In the Map tab, in the Hysteresis subtab, sort the Hysteresis column from high to low hysteresis.

3. Click Display perturbation properties and select Predicted ddG (raw) and Predicted ddG (corrected) to display the values on the edges of the graph. The raw values (blue) are the Bennett values.

4. Click Display perturbation properties and select Highlight bad perturbations. The unconverged perturbations are colored red. The algorithm used to detect these unconverged perturbations is:

   • Sort all closed cycles by hysteresis error.
   • Select the first row in the Hysteresis tab.
   • Identify the specific edge in that cycle where the difference between the raw and the predicted dG value is largest (approx. >0.5)
   • Delete this edge to see if many rows that were colored red in the Hysteresis tab disappear
   • Repeat until there are no cycles colored red, or there is no edge with a difference between the raw and predicted dG values greater than 0.5.

5. In the Analysis tab, select an unconverged edge in the table, and look at the plots in the Convergence plots for the solvent and complex legs.