[AMBER] Thermodynamic integration for charge-changing mutations

From: Skanda Sastry via AMBER <amber.ambermd.org>
Date: Thu, 5 Jan 2023 17:22:18 -0800

Dear Amber mailing list,

I hope you all had a refreshing holiday break. I wanted to write to ask
help with using thermodynamic integration to find the change in binding
free energy upon a charge-changing mutation. We are specifically using this
approach to try to predict binding free energy changes in an
antibody-antigen complex (PDB: 6OGE).

According to the literature, there are problems with the accuracy of
thermodynamic integration when the transformation does not conserve net
charge (source <https://pubs.acs.org/doi/pdf/10.1021/acs.jcim.0c00613>). To
mitigate this issue, we have been trying to implement the neutral
co-alchemical ion approach given in this paper
<https://pubs.acs.org/doi/pdf/10.1021/acs.jctc.8b00825>. However, we have
run into a few challenges with this approach.

Our process is to use 12-lambdas, 5 ns at each lambda, and then use the
average dV/dL from each lambda and a weighted sum with Gaussian quadrature
to compute the change in free energy upon a mutation. We prepare the
structure by doing minimization, heating (with backbone restrained), NPT
equilibration (with backbone restrained), and NVT equilibration
(unrestrained) at lambda = 0.5. Then, we equilibrate the system at each of
the different lambdas before doing the 5 ns production run. We run
simulations in explicit water, with 150 mM NaCl - this paper
<https://pubs.acs.org/doi/pdf/10.1021/acs.jctc.8b00825> discusses the
improvement in accuracy when running simulations in a solution with 150 mM
NaCl vs no salt.

The strange phenomenon I wanted to write in about was that in many
charge-changing simulations, we see the co-alchemical neutral ion
inevitably nearing closer and closer to the amino acid getting mutated
during the preparatory heating and NPT/NVT equilibrations. The neutral ion
is part of the TI region (timask), but not part of the SC region (scmask),
while the original and mutated amino acid are part of both the TI and SC
regions. We were wondering what causes the neutral ion to approach closer
and closer to the mutated amino acid? Currently, the only way we can keep
the neutral ion in the bulk solution is by placing a positional restraint
on the neutral ion. We feel that the neutral ion should not be close to the
solute at all, since its only purpose is for the system to remain charge
neutral during thermodynamic integration.

I have attached input files containing the parameters for minimization,
heating, and NPT/NVT equilibration in order to give as much information as
possible. I also have attached the topology and coordinates files for the




AMBER mailing list

Received on Thu Jan 05 2023 - 17:30:02 PST
Custom Search