Understanding the options of atom mapping and ligandHybrid

Dear users and developers,
How do atom mapping options for ligand RBFE differ from each other as in where I use (–H2Heavy option, whether hydrogen be morphed into a heavy atom) and I have been unable to get forward and backward convergence in my FEP calculations.

Whereas, when I use the default options and do not pass any option explicitly, I see many atoms being filled both in A and B state with DUM in the prefix which however seem to be virtual sites/dummy atoms. And in the latter case, my simulations seem to have pretty well backward and forward convergence.

And from my understanding with topologies used in amino acid FEP is that, when I am converting a Halogen atom to a Hydrogen, the hybrid topology should have been pretty straight forward rather than using such dummy atoms/virtual sites. However in such case, the aforementioned problem of forward and backward convergence.

The other query I had about is scaling of dummy masses(–scDUMm), dummy angle parameters( --scDUMa) or dihedral parameters(–scDUMd). What is it that this option essentially does?

It would be kind if anyone could explain me the nuances behind it.
P.S: I have been using BAR and equilibrium free energy calculations.
Warm regards,
Pallav Sengupta,
Guwahati, Assam, India

H2Heavy will allow mapping hydrogen atoms in one ligand to heavy atoms in another ligand. I recommend not to use this option.

Many dummies are added when the two ligands share little similarity, yet it does not necessarily mean that the convergence will suffer. As you noticed, the convergence is better with the default mapping options.

These options scale down dummy masses, angle and dihedral force constants between real and dummy atoms. This might be required to ensure simulation stability and to prevent dummy atoms from influencing internal geometry of the rest of the molecule.

1 Like

Thanks a lot Vytas.
And there is another query I have regarding soft core potentials during transformations,
how does one determine the value of sc_alpha and sc_sigma, and does these values influence efficient transformation and also affect convergence?

The soft-core parameters are determined from empirical observations to give the smooth alchemical transition and allow for good convergence. There are some standard values that have been benchmarked to behave well. You can look into the Beutler et al, 1994 paper, where several sets of softcore params were evaluated.

1 Like