Dear Anthony,

I am not an expert in the determination of free energies for these types of reactions. However, to avoid underestimation of the error of your observables you can try and use the so-called blocking method. You can read about it in the original paper:

Flyvbjerg,H.;Petersen,H. J.Chem.Phys.1989,91,88–103.

Have also a look at a more recent article:

Romo, T. D. & Grossfield, A. Block covariance overlap method and convergence in molecular dynamics simulation. J. Chem. Theory Comput. 7, 2464–2472 (2011).

Good luck,

Alessandro

Dear Anthony,

Giving a bias equal to the negative estimate of the free energy sounds more like Thermodynamic Integration with Adaptive Biasing Force sampling technique. This technique is implemented in the NAMD program package, however, and I am not sure if you could use the same force field. What seems complicated in this case, is that you need to set a proper force constant (since you apply an umbrella potential for each window), which should be estimated from the PMF values for the respective interval of your reaction coordinate.

I would recommend to look closely at the intermolecular structure and the mutual orientation of your reactants in each window. If there is a distinctive change of these two characteristics when you bring the reactants together, then you certainly have to account for transverse degrees of freedom. Generally, this is an issue, which is not easy to overcome, and some techniques such as Metadynamics and ABF could be used to validate your results from the Umbrella Sampling simulations. Otherwise, one could also use Perturbation methods.

Good luck,

Tsvetan